UBICC Journal

Ubiquitous Computing and Communication Journal

Volume 4 · Number 3 · July 2009 · ISSN 1992-8424

Special Issue on ICIT 2009 Conference – Applied Computing

UBICC Publishers © 2009

Ubiquitous Computing and Communication Journal
Co-Editor Dr. AL-Dahoud Ali

Ubiquitous Computing and

Communication Journal

Book: 2009 Volume 4

Publishing Date: 07-30-2009
Proceedings
ISSN 1992-8424

This work is subjected to copyright. All rights are reserved whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, re-use of illusions, recitation, broadcasting,
reproduction on microfilms or in any other way, and storage in data banks. Duplication of this publication of
parts thereof is permitted only under the provision of the copyright law 1965, in its current version, and
permission of use must always be obtained from UBICC Publishers. Violations are liable to prosecution under
the copy right law.

UBICC Journal is a part of UBICC Publishers

www.ubicc.org

© UBICC Journal
Printed in South Korea

Typesetting: Camera-ready by author, data conversation by UBICC Publishing Services, South Korea

UBICC Publishers

2
 Guest Editor’s Biography
Dr. Al-Dahoud, is a associated professor at Al-Zaytoonah University, Amman, Jordan.
He took his PhD from La Sabianza1/Italy and Kiev Polytechnic/Ukraine, on 1996.
He worked at Al-Zaytoonah University since 1996 until now. He worked as visiting professor in many
universities in Jordan and Middle East, as supervisor of master and PhD degrees in computer science. He
established the ICIT since 2003 and he is the program chair of ICIT until now. He was the Vice President of the
IT committee in the ministry of youth/Jordan, 2005, 2006. Al-Dahoud was the General Chair of (ICITST-2008),
June 23–28, 2008, Dublin, Ireland (www.icitst.org).
He has directed and led many projects sponsored by NUFFIC/Netherlands:
- The Tailor-made Training 2007 and On-Line Learning & Learning in an Integrated Virtual Environment" 2008.
His hobby is conference organization, so he participates in the following conferences as general chair, program
chair, session’s organizer or in the publicity committee:
- ICITs, ICITST, ICITNS, DepCos, ICTA, ACITs, IMCL, WSEAS, and AICCSA
Journals Activities: Al-Dahoud worked as Editor in Chief or guest editor or in the Editorial board of the
following Journals:
Journal of Digital Information Management, IAJIT, Journal of Computer Science, Int. J. Internet Technology and
Secured Transactions, and UBICC.
He published many books and journal papers, and participated as speaker in many conferences worldwide.
 UBICC Journal
Volume 4, Number 3, July 2009

SPECIAL ISSUE ON ICIT 2009 CONFERENCE: APPLIED COMPUTING

538 The influence of blended learning model on developing leadership skills

of school administrators
Tufan Aytaç

544 Towards the implementation of temporal-based software version

management in Universiti Darul Iman Malaysia
Mohd Nordin Abdul Rahman, Azrul Amri Jamal, Wan Dagang Wan Ali

551 Effective digital forensic analysis of the NTFS disk image

Mamoun Alazab, Sitalakshmi Venkatraman, Paul Watters

559 Job and application-level scheduling in distributed computing

Victor V. Toporkov

571 Least and greatest fixed points of a while semantics function

Fairouz Tchier

585 Case studies in thin client acceptance

Paul Doyle, Mark Deegan, David Markey, Rose Tinabo, Bossi Masamila, David Tracey

599 An interactive composition of workflow applications based on UML

activity diagram
Vousra Bendaly Hlaoui, Leila Jemni Ben Ayed

609 How to map perspectives

Gilbert Ahamer

4
 Special Issue on ICIT 2009 Conference - Applied Computing

THE INFLUENCE OF BLENDED LEARNING MODEL ON DEVELOPING

LEADERSHIP SKILLS OF SCHOOL ADMINISTRATORS

Tufan AYTAÇ
The Ministry of National Education, Ankara, TURKEY
taytac1@yahoo.com

ABSTRACT

The usage of b-learning approach on in-service education activities in Turkish education system are
getting more and more important these days. Generally, traditional education and computer based
education applications are used on in-service education activities. Blended learning (b-learning)
combines online learning with face-to-face learning. The goal of blended learning is to provide the most
efficient and effective learning experience by combining learning environments. The purpose of this
research is to find out the effect of b-learning approach on developing administrators’ leadership skills.
To identify what the school administrators’ educational needs and to know their existing leadership
skills, needs assessment questionnaire was applied to 72 school administrators who were selected from
33 primary schools in 11 region of Ankara capital city. According to the descriptive statistical analysis
results of questionnaire, in-service training programme was prepared for the development of school
administrators’ leadership skills. The school administrators were separated into three groups as
computer based learning (CBL) (25 participants), blended learning (BL) (23 participants) and
traditional learning (TL) (24 participant) groups. These groups were trained separately with these three
different learning environments by using the in-service training programme. According to the results of
pre-test, post test and achievements score means, it was observed that BL groups’ score is the highest
when compared to TL and CBL groups. As a result of this research, in terms of achievements and
effectiveness, b-learning was found to be the most effective learning environment when compared to the
others. Both learners and tutors findings strongly suggest that blended learning is available alternative
delivery method for inservice education activities.1

Keywords: Blended Learning, e-Learning, Information Technology, In-service education

innovative and technological advances offered by online

1 INTRODUCTION
Blended Learning (b-Learning or Hybrid Learning) consists
of the combination of e-Learning and traditional education
approach. Blended learning combines online learning with
face-to-face learning. The goal of blended learning is to
provide the most efficient and effective learning experience
by combining different learning environments. b-Learning
stands in the forefront in respect of interactivity with target
learner group, enriching learning process and integration of
technology into education [1,2,3,16,21].
E-learning has had an interesting impact on the learning
environment. Blended learning is the most logical and natural
evolution of our learning agenda. It suggests an elegant
solution to the challenges of tailoring learning and
development. It represents an opportunity to integrate the
learning with the interaction and participation offered in the
best of the traditional learning [20].
The ground of blended learning approach constitutes the
powerfull side of traditional education and computer based
educations instead of using one or the other on its own.
Basic characteristics of Blended learning which reflects
the values of 21st century education are [2];
 Providing a new way of learning and teaching,
 Teaching how to learn,
 Creating digital learners,
 Be more economical,
 Focusing on technology and communication
 Improving project-based learning,
 And improving teaching process.

1
This research project article has been supported by The Scientific and Technological Research Council of Turkey (TÜBİTAK) (SOBAG 1001
Programme).

UbiCC Journal – Volume 4 No. 3 538

 Special Issue on ICIT 2009 Conference - Applied Computing
Blended Learning practices provide project based learning
opportunities for active learning and interaction among
learners and especially provides as a way to meet the
educational needs of the learners. Blended learning programs
may include several forms of learning tools, such as real-time
virtual/collaboration software, self-paced web-based courses,
electronic performance support systems (EPSS) embedded
within the learning-task environment, and knowledge
management systems. Blended learning contains various
event-based activities, including face-to-face learning, e-
learning, and self-paced learning activities. Blended learning
often occurs as a mixture of traditional instructor-led training,
synchronous online training, asynchronous self-paced study,
and structured task based training from a teacher or mentor.
The aim of blended learning is to combine the best of
classroom face-to-face learning experiences with the best of
online learning experiences. Overall, blended learning refers
to the integration (or the so-called blending) of e-learning
tools and techniques with traditional face-to-face teaching
delivery methods. The two important factors here are the time
spent on online activities and the amount of technology
utilized, see Concept of Blended Learning figure 1 below:
[3,4,6,7,8,9,10,11,12,15,16,19].
Fig. 1: Concept of Blended Learning
If two or more of these learning environments which are
stated above are used to teach an educational objective, it can
be said that blended learning is realized. However blended
learning has more meaning than showing a web page during
a lesson in the classroom and using information immediately
in the web page to explain the lesson. Blended learning is a
learning of environment which combines environments of
face to face learning and web-based distance learning.
Blended learning overcomes this limitation of an e-
learning only approach [12]. Today blended learning
primarily functions as a replacement for extension of face-to
face environments. For instance, it might be used to foster
learning communities, extend training events, offer follow-up
resources in a community of practice, access guest experts,
provide timely mentoring or coaching, present online lab or
simulation activities, and deliver prework or supplemental
UbiCC Journal – Volume 4 No. 3
course materials. While such uses may be unique and
engaging, they are not exactly novel [13].
Figure 2: A Blend of Learning Theories
By applying learning theories of Keller, Gagné, Bloom,
Merrill, Clark and Gery, (see Figure 2) five key ingredients
emerge as important elements of a blended learning process
(see Figure 2):
1. Live Events: Synchronous, instructor-led learning events
in which all learners participate at the same time, such as in a
live “virtual classroom.”
2. Self-Paced Learning: Learning experiences that the
learner completes individually, at his own speed and on his
own time, such as interactive, Internet-based or CD-ROM
training.
3. Collaboration: Environments in which learners
communicate with others, for example, e-mail, threaded
discussions or online chat.
4. Assessment: A measure of learners’ knowledge. Pre-
assessments can come before live or self-paced events, to
determine prior knowledge, and post-assessments can occur
following live or self-paced learning events, to measure
learning transfer.
5. Performance Support Materials: On-the-job reference
materials that enhance learning retention and transfer,
including PDA downloads, and printable references,
summaries, and job aids.
2 PURPOSE
The purpose of this research is to find out the effects of
b-learning approach on developing school administrators’
leadership skills.
3 RESEARCH DESIGN
To determine what the school administrators’ educational
needs on leadership skills, needs assessment questionnaire
was applied to 72 school administrators who were selected
from 33 primary schools in 11 regions within Ankara capital
city. According to the results of this questionnaire, in-service
539
 Special Issue on ICIT 2009 Conference - Applied Computing

training programme on developing school administrator’s

leadership skill was prepared.
The most needed leadership skills of school administrators
according to the results of needs assessment were determined
as human relations in administration, basic management
skills for school principles, job satisfaction at organizations
and motivation.
After that, content and learning activities of "School
Administrators Leadership Skills Development In-service
Programme" were prepared. Beside that course notes as
training materials were prepared to be distributed to the
participants in the form of CDROM and printed documents.
The school administrators were separated into three groups
as Computer Based Learning (CBL) (25 participants),
Blended Learning (BL) (23 participant) and Traditional
Learning (TL) (24 participant) groups. These groups were
trained according to three different methods by preparing
education programme. The groups were given two days
course.
Before the in-service training the school administrators
who were in BL group reached the digital content and studied
learning activities included in "School Administrators
Leadership Skills Development In-service Programme" which
is prepared by using Moodle Learning Managing System
Figure. 3: The Moodle interface
Softwware and published on http://beg.meb.gov.tr:8088/
website. On the other hand; all the in-service training content and
The school administrators who are in the BL group were activities were taught to CBL group by lecturer with aid of
entered to the http://beg.meb.gov.tr:8088/ webpage by using computer and projector. Finally, TL group was trained in a
their usernames and passwords given to them three weeks traditional way by using blackboard
ago, before the in-service training. The interface of the Multiple choice test which was made up of 20 questions
website is shown in the Fig. 2. The school administrators in were applied to the groups to investigate their achievements
this group shared information, chatted, and studied activities on leadership skills. This test was shown to content experts to
with their colleagues and subject area specialist about the identify its content validity. To find out the statistical
content and learning activities included in the site whenever significant difference among three groups score means, one-
they want. As online learner, school administrators build way Anova and Scheffe test were used. This test was applied
their confidence and learning processes as they get used to to all groups as pre-test at the beginning and as post-test at
working independently online. Blended learning activities the end of in-service training [5]. Blended Learning Model
included online knowledge gathering and construction in which was used on the research process showed Figure 3.
teams or groups, publishing of electronic content, interactive
elements like online brainstorming, discussion, several forms
of feedback, evaluation and assessment, as well as other
blended learning techniques. Lecturers posted messages to the
BL group as a whole and to each administrators individually
to meet their need for support. They posted explanation to
guide learners in more complex tasks, encouraged them to
communicate, to do their individual assignments, and to use
the Moddle platform tools. They have at their disposal to
facilitate their work. Tutors controlled and marked the online
assignments, filled in learners’ performance reports, and
write feedback on their performance in their online portfolios.
Lecturers followed school administrators learning
improvements and gave encouragement when motivation
level began to falter. And after that this group was trained by
lecturer as subject area specialist. Lecturer trained this group Figure. 3: The Process of Blended Learning Model
by using face to face education, computer based education and
online training website prepared by moodle software.

UbiCC Journal – Volume 4 No. 3 540

 Special Issue on ICIT 2009 Conference - Applied Computing

4 RESULTS Table 3: The One-Way Anova Results on the difference

Between Groups According to Achievement Scores
When three groups’ pre-test score means were
(Difference between pre-test and post test)
compared, it was seen that there were significant differences
among them (F (2-69)=53,350, p<.01). (Table I).
Source of Sum of df Mean Mean
Table 1: The One-Way Anova Results on the Difference Variance Squares Square F Sig. Difference
Between Groups According to Pre-Test Scores
Between BDE-KÖ,
46,540 2 23,270 18,086 ,000
groups GÖ-KÖ
Source of Sum of df Mean Mean
Variance Squares Square F Sig. Difference Within
88,779 69 1,287
groups
Between BDE-KÖ,
278,668 2 139,334 53,350 ,000 Total 135,319 71
groups GÖ-KÖ
Within It was seen that BL group’s pre-test, post test and
180,207 69 2,612 achievements score means were the highest when compared
groups
to TL and CBL groups. The reason of this might be that BL
Total 458,875 71
group might be more ready than others since they studied
When three groups were statistically compared content and activities which were published with Moodle
according to data, BL groups’ pre-test score mean ( =12.87) software before other groups. They also experienced both face
was found statistically higher than the other two groups (BDE to face and computer based learning environments.
=9.12, TL =8.29). The reason of this might be that BL
group was more ready and successful than others. Since they
5 CONCLUSION
studied earlier all content and activities which were prepared
with Moodle software and published on internet. The influence of b-learning model on developing
leadership skills of school administrators was more effective
Table 2: The One-Way Anova Results on the Difference than computer based education and traditional learning.
Between Groups According to Post-Test Scores As a result of this research, in terms of time, cost and
effectiveness, b-Learning was found to be the most effective
Source of Sum of df Mean Mean method to in comparison with the other approaches.
Variance Squares Square Difference Particularly, it appeared that it is necessary to use more b-
F Sig.
Learning approach in in-service training of administrators
Between BDE-KÖ, and teachers. It is required effective usage of b-Learning
90,61
groups 544,539 2 272,270 ,000 BDÖ-GÖ, approaches for integrating education with information
0
GÖ-KÖ technologies, enriching learning-teaching process,
Within implementing face to face education, providing computer
207,336 69 3,005
groups based learning, realizing hands on learning and
Total 751,875 71 individualizing the learning.
When school administrators’ post test score means were At the research blended-learning arrangements involved
compared among three groups, it was found that there was a e-mentoring or e-tutoring. The role of the e-mentor/tutor is
significant difference between post test score means. (F (2- critical as this requires a transformation process to that of
69)=90,610, p<.01). (Table II). There was also statistically learning facilitator. Being teachers and online tutors has
significant differences among BL group’s post test score introduced beneficial qualitative changes in teachers’ roles,
mean ( =17.35), CBL group’s post test score mean ( =12.44) but it has also meant a quantitative increase in the number of
and TL group’s post test mean ( =10.79). Especially, BL hours dedicated to learners. Lecturers less spent time in face-
group administrators’ post test score mean is the highest of to-face classes than the online environment (Moddle
all. platform).
The difference between school administrators’ pre-test Moodle programme which is used blended learning
and post test scores was being calculated to identify their approach has great potential to create a successful blended
achievement scores. It was seen that there was a meaningful learning experience by providing a plethora of excellent tools
difference among the groups’ achievement score means (F (2- that can be used to enhance conventional classroom
69)=18.086, p<.01) (Table 3). BL group’s achievement instruction, in hybrid courses, or any distance learning
means ( =4.48) has been higher than CBL ( =3.28) and TL arrangements [18].
( =2.50) groups’ achievement means. Finally, lecturers identified learners who may be
experiencing particular problems and help them address their
weakneses in remedial work sessions if necessary.

UbiCC Journal – Volume 4 No. 3 541

 Special Issue on ICIT 2009 Conference - Applied Computing
We observed that b-Learning opportunities for teaching
objectives make learning entertaining, funny, lasting and
economics as an effective way. In this sense, according to us
trainers should use b-Learning environment for the
integration of ICT effectively in learning and teaching.
Last year, the Turkish Ministry of National Education In-
service Training Department implemented more than 700 in-
service training courses. The usage of b-Learning
methodology especially in these in-service trainings will
enrich and support the learning-teaching process of those in-
service training. More projects about the usage of b-Learning
in-service training should be supported and performed.
Particularly, the initiatives of the Turkish Ministry of
National Education for improving schools information
technologies and internet infrastructure, distributing
authoring software to the teachers, developing education
portal and its content, moodle and similar learning
management system software should be used for supporting b-
learning usage in-service training. School administrators
state that b-learning approaches will be used more effectively
in the class. All school administrators’ comments regarding
the blended course were positive.
It is cited as below that the positives of the blended learning
course activities which are used at this research;
- Improvement in the quantity and/or quality of the
communications among the school administrators in
discussion board or online groups and face to face activities in
the classroom.
- Good cooperative learning activities
- Blended learning were more effective than classroom
alone. Higher learner value and impact; the effectiveness
greater than for nonblended approaches. Learners like b-
learning appraoches.
- Accessibility to b-learning content and activities rapidly
(every time, everywhere)
- Improved relationships between tutors and students
- The immediate feedback that could be given school
administrators
- Flexibility in scheduling and timetabling of course work.
- An increase of the time actually spent on face-to-face in
classroom
- Cost effectiveness for both the accrediting learning
institution and the learner
The increased cost, reduced training time, and the ability to
easily update training materials offer additional compelling
reasons for educators to embrace blended learning [22].
At the research there are some problems according to school
administrators opinions cited as below:
- Some technical web, internet problems access to moddle
platform.
- The failure of online Power Point presentation of lecture
material to meet some school administrators’ expectations.
- Some school administrations lack of enthusiasm being in a
blended learning course.
- Limited knowledge in the use of technology.
UbiCC Journal – Volume 4 No. 3
- Blended learning takes time for both the instructor and
learner to adapt to this relatively new concept in delivering
instruction.
Especially, it can be concluded that all the in-service
training should be taught more effectively by using b-
Learning approach. The technological leadership role of the
school administrations is very important for the success of b-
Learning approach.
The feature of blended learning models has a vital
importance for applying individual learning and active
learning. According to some authors “a blend is integrated
strategy to delivering on promises about learning and
performance [17].
In sum, both learners and tutors findings strongly suggest
that blended learning is available alternative delivery method
for courses. In supporting blended learning, especially in-
service education courses remains both a national leader in
the effective use of technology for teaching and learning, and
a pioneer in identifying the right mix of face-to-face and
online communication practices that will enhance learning
effectiveness [19]. The result of this research backs up all of
these. To develop the technological leadership of school
administrators, b-learning approaches should be used
effectively. Blended learning offers opportunities for both in-
service school administrators, in-service teachers and their
learners.
REFERENCES
[1] Aytaç, T. Eğitimde Bilişim Teknolojileri. Asil Yayın
Dağıtım, pp. 48-53 (2006).
[2] Aytaç, T. The Influence of B-Learning Model on
Developing Leadership Skills of Education
Administrators Research Education Programme, pp.
48-53. (2006).
[3] Singh, H. “Building Effective Blended Learning
Programs”, Educational Technology, Vol. 43, Number
6, pp. 51-54, November – December, (2003).
[4] Oliver, M. ve Trigwell, K. ‘‘Can ‘Blended Learning’ Be
Redeemed?. E-Learning, Vol.2. Number 1, pp. 17,
(2005).
[5] Büyüköztürk, Ş. Sosyal Bilimler İçin Veri Analizi El
Kitabı. İstatistik, Araştırma deseni SPSS
Uygulamaları ve Yorum, 8. Baskı, PegemA Yayıncılık,
Pp: 40-53, Ankara, (2007).
[6] Bonk, C. J.; Olson, T. M.; Wisher, R. A. and Orvis, K. L.
Learning from Focus Groups: An Examination of
Blended Learning’’, Journal of Distance Education.
Vol. 17, No 3. pp. 100. (2002).
[7] Marsh, J. How to Design Effective Blended Learning.
www.brandon-hall.com. Erişim tarihi: 15 February 2009.
[8] Orhan, F. Altınışık, S. A. and Kablan, Z. “Karma
Öğrenme (Blended Learning) Ortamına Dayalı Bir
542
 Special Issue on ICIT 2009 Conference - Applied Computing

Uygulama: Yıldız Teknik Üniversitesi Örneği”, IV. [20] Thorne, K. Blended Learning: How to Integrate
Uluslararası Eğitim Teknolojileri Sempozyumu, 24-26 Online and Traditional Learning. United States, Kogan
Kasım 2004, Sakarya, Vol: 1, pp.646-651, (2004). Page, (2004).
[21] Rovai, Alfred P. and Jordan, Hope M. "Blended Learning
[9] Dracup, Mary. "Role Play in Blended Learning: A Case
with Traditional and Fully Online Graduate Courses."
Study Exploring the Impact of Story and Other Elements,
International Review of Research in Open and
Australasian Journal of Educational Technology,
Distance Learning. 2004. Retrieved Sept 27, from
24(3), pp.294-310, (2008).
http://www.irrodl.org/content/v5.2/rovaijordan.html.
[10] Cooper, G. and Heinze, A. "Centralization of (2008)
Assessment: meeting the challenges of Multi-year Team [22] G. Thorsteinsson and T. Page. “Blended Learning
Projects in Information Systems Education." Journal of Approach to Improve, In-Servıce Teacher Education In
Information Systems Education, 18, 3, pp.345 – 356, Europe Through The Fıste Comenıus 2.1. Project” Ict in
(2007). Education: Reflections and Perspectives, Bucharest,
June 14-16, (2007).
[11] Heinze, A. Lecturer in Information Systems,
http://www.aheinze.me.uk/Blended_Learning_Higher_E
ducation.html, Erişim tarihi: 15 February 2009.
[12] Langley, Amanda. “Experiential Learning, E-Learning
and Social Learning: The EES Approach to Developing
Blended Learning” The Fourth Education in a
Changing Environment Conference Book, Edited by
Eamon O’Doherty, İnforming Science Press, pp.171-172,
(2007).
[13] Bonk, C. J. & Graham, C. R. (Eds.). “Future Directions
of Blended Learning In Higher Education and
Workplace Learning Settings” Handbook of blended
learning: Global Perspectives, local designs. San
Francisco, CA: Pfeiffer Publishing. (2004).
[14] Carman, Jared M. Blended Learning Design: Five Key
Ingredients, Director, Product Development
KnowledgeNet, October 2002 www.brandon-hall.com.
Erişim tarihi: 15 February 2009.
[15] Derntl M. Motschnig-Pitrik, Renate. A Layered
Blended Learning Systems Structure, Proceedings of I-
KNOW ’04 Graz, Austria, June 30 - July 2, (2004).
[16] Bañadosa, Emerita. Blended-learning Pedagogical
Model for Teaching and Learning EFL Successfully
Through an Online Interactive Multimedia Environment,
CALICO Journal, Vol. 23, No. 3, p-p 533-550, (2006).
[17] Rosset, A., Douglis, F. &Frazee, R. V. Strategies for
building blended learning. Learning Circuits.
Retrieved August 13, 2007, from
http://www.learningcircuits.org/2003/jul2003/rossett.htm
.
[18] Brandl, K. (2005). Are you ready to moodle?. Language,
Learning & Technology, Vol. 9, No. 2, pp. 16-23, May
(2005).
[19] Blended Learning Pilot Project, Final Report for 2003-
2004 and 2004-2005 Rochester Institute of Technology.
(2004). Blended Learning Pilot Project: Final Report
for the Academic Year 2003 – 2004. Retrieved Feb 5,
,fromhttp://distancelearning.rit.edu/blended/Files/Blende
dPilotFinalReport2003_04.pdf. (2009).

UbiCC Journal – Volume 4 No. 3 543

Special Issue on ICIT 2009 Conference - Applied Computing

TOWARDS THE IMPLEMENTATION OF TEMPORAL-BASED

SOFTWARE VERSION MANAGEMENT AT UNIVERSITI DARUL IMAN
MALAYSIA

M Nordin A Rahman, Azrul Amri Jamal and W Dagang W Ali

Faculty of Informatics
Universiti Darul Iman Malaysia, KUSZA Campus
21300 K Terengganu, Malaysia
mohdnabd@udm.edu.my, azrulamri@udm.edu.my, wan@udm.edu.my

ABSTRACT
Integrated software is very important for the university to manage day-to-day operations. This integrated software is
going through evolution process when changes are requested by the users and finally the new versions are created.
Software version management is the process of identifying and keeping track of different versions of software.
Complexity level of this process would become complicated should software was distributed in many places. This
paper presents a temporal-based software version management model. The model is purposely implemented for
managing software versions in Information Technology Centre, Universiti Darul Iman Malaysia. Temporal elements
such as valid time and transaction time are the main attributes considered, to be inserted into the software version
management database. By having these two attributes, it would help the people involved in software process to
organize data and perform monitoring activities with more efficient.

Keywords: version management, temporal database, valid time, transaction time.

1. INTRODUCTION  The use of data to conduct analysis of past events

Software evolution is concerned with modifying
software once it is delivered to a customer. Software
managers must devise a systematic procedure to
ensure that different software versions may be
retrieved when required and are not accidentally
changed. Controlling the development of different
software versions can be a complex task, even for a
single author to handle. This task is likely to become
more complex as the number of software authors
increases, and more complex still if those software
authors are distributed geographically with only
limited means of communication, such as electronic
mail, to connect them.
Temporal based data management has been a hot
topic in the database research community since the
last couple of decades. Due to this effort, a large
infrastructure such as data models, query languages
and index structures has been developed for the
management of data involving time [11]. Nowadays,
a number of software has adopted the concepts of
temporal database management such as artificial
intelligence software, geographic information systems
and robotics. Temporal management aspects of any
objects could include:
 The capability to detect change such as the
amount of change in a specific project or object
over a certain period of time.
e.g., the change of valid time for the project or
version due to any event.
 To keep track of all the transactions status on the
project or object life cycle.
Universiti Darul Iman Malaysia (UDM) is the
first full university at East Cost of Malaysia located at
the state of Terengganu. It was setup on 1st January
2006. UDM has two campus named as KUSZA
Campus and City Campus. Another new campus
known as Besut Campus will be operated soon. To
date, KUSZA Campus has six faculties and City
Campus has three faculties. The university also has
an Information Technology Centre (ITC-UDM) that
purposely for developing and maintaining the
university information systems and information
technology infrastructure.
In this paper, we concentrate on the modelling of
a temporal-based software version management.
Based on the model, a simple web-based web
application has been developed and suggested to be
used by ITC-UDM. The rest of the paper is organized
as follows: next section reviews the concept of
temporal data management. Section 3 discusses on
the current techniques in software version
management. Current issues in software version
management at ITC-UDM are discussed in Section 4.
The specifications of the proposed temporal-based
software version management model are explained in
Section 5. Conclusion is placed in Section 6.

UbiCC Journal – Volume 4 No. 3 544

Special Issue on ICIT 2009 Conference - Applied Computing
2. TEMPORAL DATA CONCEPT
To date, transaction time and valid time are the two
well-known of time that are usually considered in the
literature of temporal database management [2, 4, 6,
9, 10, 11, 12]. The valid time of a database fact is the
time when the fact is true in the miniworld [2, 6, 9,
10]. In other words, valid time concerns the
evaluation of data with respect to the application
reality that data describe. Valid time can be
represented with single chronon identifiers (e.g.,
event time-stamps), with intervals (e.g., as interval
time-stamps), or as valid time elements, which are
finite sets of intervals [9]. Meanwhile, the transaction
time of a database fact is the time when the fact is
current in the database and may be retrieved [2, 6, 9,
10]. This means, that the transaction time is the
evaluation time of data with respect to the system
where data are stored. Supporting transaction time is
necessary when one would like to roll back the state
of the database to a previous point in the time. [9]
proposed four implicit times could be taken out from
valid time and transaction time:
 valid time – valid-from and valid-to
 transaction time – transaction-start and
transaction-stop
Temporal information can be classified into two
divisions; absolute temporal and relative temporal [9].
Most of the research in temporal databases
concentrated on temporal models with absolute
temporal information. To extend the scope of
temporal dimension, [12] presented a model which
allows relative temporal information e.g., “event A
happened before event B and after January 01, 2003”.
[12] suggests several temporal operators that could be
used for describing the relative temporal information:
{equal, before, after, meets, overlaps, starts, during,
finishes, finished-by, contains, started-by, overlapped-
by, met-by and after}.
In various temporal research papers the theory of
time-element can be divided into two categories:
intervals and points [6, 9, 11]. If T is denoted a
nonempty set of time-elements and d is denoted a
function from T to R+, the set of nonnegative real
numbers then:
interval , if d(t)  0
time _ element, t  
 point , otherwise
According to this classification, the set of time-
elements, T, may be expressed as T = I  P, where I
is the set of intervals and P is the set of points.
UbiCC Journal – Volume 4 No. 3
3. RELATED TOOLS IN SOFTWARE VERSION
MANAGEMENT
In distributed software process, a good version
management combines systematic procedures and
automate tools to manage different versions in many
locations. Most of the methods of version naming use
a numeric structure [5]. Identifying versions of the
system appears to be straightforward. The first
version and release of a system is simply called 1.0,
subsequent versions are 1.1, 1.2 and so on.
Meanwhile, [3] suggests that every new version
produced should be placed in a different directory or
location from the old version. Therefore, the version
accessing process would be easier and effective.
Besides that, should this method be implemented
using a suitable database management system, the
concept of lock access could be used to prevent the
occurrence of overlapping process. Present, there are
many software evolution management tools available
in market. Selected tools are described as follows:
 Software Release Manager (SRM) – SRM is a
free software and supported on most UNIX and
LINUX platforms. It supports the software
version management for distributed
organizations. In particular, SRM tracks
dependency information to automate and
optimize the retrieval of systems components as
well as versions.
 Revision Control System (RCS) – RCS uses the
concepts of tree structures. Each branch in the
tree represents a variant of the version. These
branches will be numbered by an entering
sequence into a system database. RCS records
details of any transaction made such as the
author, date and reason for the updating.
 Change and Configuration Control (CCC) –
CCC is one of the complete tools for software
configuration management. It provides a good
platform for an identification, change control and
status accounting. CCC allows a simultaneously
working for a same version via virtual copies.
This can be merged and changes can be applied
across configurations.
 Software Management System (SMS) – SMS
allows all the aspects in software configuration
management such as version control, workspace
management, system modelling, derived object
management, change detection in the repository
etc. SMS possesses the desired characteristics,
providing resources of version control of systems
and having a good user interface.
545
Special Issue on ICIT 2009 Conference - Applied Computing
4. THE SOFTWARE VERSION MANAGEMENT
ISSUES IN ITC-UDM
There are three divisions have been formed at ITC-
UDM. These divisions and their function are as
follows:
 Infrastructure and Application Systems (AIS) –
to develop and maintain the university software;
maintain the university computer networking;
 Technical and Services (TS) – to support the
maintenance of information technology
hardware, training, multimedia services and help
desk.
 Administration and Procurement (AP) - to
manage the daily operation of ITC-UDM such as
administration, procurement etc.
Each division is headed by a division leader and
supported by several information technology officers,
assistant information technology officers and
technicians. All the university software modules are
developed and maintained by AIS Division. Figure 1
depicts the main software modules managed by the
ITC-UDM. There are over thousands source code
files are produced by the division. Therefore, it is not
easy for the division to manage all those artefacts.
Academic Module
University Software Module
Human Resource Module
Student Affairs Module
Finance Module
Department of Development
Figure 1: University Software Modules
From study done by the authors, two main
weaknesses have been found in the current approach
for ITC-UDM in managing all versions of source
codes produced:
 Non systematic procedure used for managing
software versions and it is difficult to recognize
the valid time for each version.
 The current approach does not consider the
aspect of relative temporal in representing the
valid time for each version.
UbiCC Journal – Volume 4 No. 3
 The current approach maintains only the concept
of current view version of which an existing
version will be overwritten by a new incoming
version during the process of an update.
Based on the mentioned problems, we strongly
believe that the development of temporal-based
software version management tool for ITC-UDM
could gain the following benefits:
 To support project and software managers in
planning, managing and evaluating version
management.
 Assigning timestamps (absolute and relative) to
each transaction will provide transaction-time
database functionality, meaning to retain all
previously current database state and making
them available for time-based queries.
 To increase the effectiveness and efficiency of
the collaborative software version management
process.
5. THE MODEL
Version control is one of the main tasks in software
configuration management. For any software version
would have its own valid time. The collection of
software versions should be organized into systematic
way for the purpose of retrieval efficiency and to
recognize valid time of those versions. Besides the
used of unique sign for the associate version, the
method of time-stamping is also needed to be
embedded into the version management database.
5.1 The Temporal-Based Version Management
Specifications
Temporal elements involved in the model are
transaction time (tt), absolute valid time (avt) and
relative valid time (rvt) which can be denoted as, TE
= {tt, avt, rvt}. Transaction time is a date-stamping
and it represents a transaction when a new valid time
for a version is recorded into the application database.
Absolute valid time is represent by two different
attributes known as valid-from and valid-until and it
also using an approach of date-stamping. Meanwhile,
relative valid time which involves a time interval, will
be represented by a combination of temporal
operators, OPERATORs = {op1, op2, op3, …, opn}
and one or more defined event(s), signed as EVENTs
= {event1, event2, event3, …, eventn}. This model,
considered only five temporal operators, hence will be
denoted as OPERATORs = {equal, before, after,
meets, met_by}. Table 1 illustrates the general
definitions of temporal operators based on time
interval and time points. Figure 2 shows the
546
Special Issue on ICIT 2009 Conference - Applied Computing

organization of temporal elements that involved in S = {A1, A2, A3, …, An, tt, avt-from, avt-until, rvt-
software version management. If we have a software from, rvt-until}
with a set of version signed as, V = {v1, v2, v3, …, vn}
then the model is: where, Ai = attribute name of a version, tt  P and,
avt-from, avt-until, rvt-from and rvt-until  T.
TEMPORAL(vi  V)  (tt  avt  rvt)
Table 2 exhibits the temporal-based version-
where, record management for representing KEWNET’s
avt = [avt-from, avt-until], software version history. For example, KEWNET
rvt = [rvt-from, rvt-until], Ver. 1.1 has been updated three times. For the first
rvt-from = {{opi  OPERATORs}  {eventi  time, the version has been recorded on tt3 with
EVENTs}} and, absolute valid time is from avf2 to avu3 and relative
rvt-until = {{opi  OPERATORs}  {eventi  valid time is from rvf2 to rvu3. For the second
EVENTs}}. updated, on tt4, absolute valid time is from avf2 to
avu4 and relative valid time is from rvf2 to rvu4. The
Thus, if the software that has a set of feature attributes version has another change request and therefore the
Ai then a complete scheme for a temporal-based in version would have a new absolute valid time from
software version management can be signed as: avf2 to avu5 and relative valid time from rvf2 to rvu5.
This transaction is recorded on tt5.

Table 1: The definitions of temporal operator base on time point and time interval

Temporal Operator Time Point Time Interval

equal t = {(t = ti)  T}  ={( = i)  T}
before  = {( < ti)  T}  = {( < i)  T}
after  = {( > ti)  T}  = {(  i)  T}
meets  = {(  ti)  T}  = {(  i)  T}
met_by  = {(  ti)  T}  = {(  i)  T}

Transaction time

Software
version

Valid time

Absolute Relative

From Until From Until

Figure 2: Temporal elements in software version management

UbiCC Journal – Volume 4 No. 3 547

Special Issue on ICIT 2009 Conference - Applied Computing

Table 2: Version-Record for KEWNET software

Ver # tt avt-from avt-until rvt-from rvt-until

1.0 tt1 avf1 avu1 rvf1 rvu1
1.0 tt2 avf1 avu2 rvf1 rvu2
1.1 tt3 avf2 avu3 rvf2 rvu3
1.1 tt4 avf2 avu4 rvf2 rvu4
1.1 tt5 avf2 avu5 rvf2 rvu5
1.2 tt6 avf3 avu6 rvf3 rvu6
1.2 tt7 avf3 avu7 rvf3 rvu7
2.0 tt8 avf4 avu8 rvf4 rvu8
2.0 tt9 avf4 avu9 rvf4 rvu9
2.1 tt10 avf5 avu10 rvf5 rvu10

5.2 The Temporal-Based Version Management
Functionality
To carry out experiments validating the model
proposed, a client-server prototype has been
developed. The prototype has three main modules:
register version, update the version valid time and
queries.
During the register version process the software
manager needs to record the foundations information
of the software version. Attributes that needed to be
key-in by software manager can be signed as, Av =
{version code, date release, version description,
origin version code, version id}. Figure 3 illustrates
the screen sample used to register the basic
information of the software version.

Figure 3: Register the software version

On completion of new software version
registration, then the software manager needs to
update its valid time and this can be done by using the
module update the version valid time, illustrated in
Figure 4. The attributes for this module formed as AT
= {version code, transaction date, description, date
start, date end, time start, time end, update by,
position}. Attribute transaction date is the current
date and will be auto-generated by the server.
Any changes of a software version valid time,
software manager needs to update by using this form.
The tool also allows the user to make a query to the
database. The users can browse the version valid time
and status for any registered software as shown in
Figure 5. Meanwhile, Figure 6 shows the output form
of query for all histories of valid time and status for a
software version.

UbiCC Journal – Volume 4 No. 3 548

Special Issue on ICIT 2009 Conference - Applied Computing

Figure 4: Update the software version valid time

Figure 5: The software version valid time report

Figure 6: The transaction records of a version

UbiCC Journal – Volume 4 No. 3 549

Special Issue on ICIT 2009 Conference - Applied Computing
6. CONCLUSION
In practical software version management, it is
frequently important to retain a perfect record of past
and current valid time for a version states. We cannot
replace or overwritten the record of old valid time of a
software version during the updating process. Hence,
this paper introduces a new model in software version
management based on temporal elements. Here, an
important issue discussed is temporal aspects such as
valid time and transaction time have been stamped on
each software version so that the monitoring and
conflict management processes can be easily made.
Based on the proposed model, a prototype has
been developed. The prototype will be experimented
in ITC-UDM. It will be used to monitor and keep
track the evolution of the software version, systems
module and software documents in university’s
software. For further improvements, currently, we are
investigating related issues including combining the
model with change request management, considering
more temporal operators and developing a standard
temporal model for all configuration items in software
configuration managements.
References:
[1] Bertino, E., Bettini, C., Ferrari, E. and Samarati,
P. “A Temporal Access Control Mechanism for
Database Systems”, IEEE Trans. On Knowledge
and Data Engineering, 8, 1996, 67 – 79.
[2] C. E. Dyreson, W. S. Evans, H. Lin and R. T.
Snodgrass, “Efficiently Supporting Temporal
Granularities”, IEEE Trans. On Knowledge and
Data Engineering, Vol. 12 (4), 2000, 568 – 587.
[3] G. M. Clemm. “Replacing Version Control With
Job Control”, ACM – Proc. 2nd Intl. Workshop
On Software Configuration Management, 1989,
162 – 169.
[4] D. Gao, C. S. Jensen, R. T. Snodgrass and M. D.
Soo, “Join Operations in Temporal Databases”,
UbiCC Journal – Volume 4 No. 3
The Very Large Database Journal, Vol. 14,
2005, 2 – 29.
[5] A. Dix, T. Rodden, and I. Sommerville.
“Modelling Versions in Collaborative Work”,
IEE – Proc. Software Engineering, 1997, 195 –
206.
[6] H. Gregerson, and C. S. Jensen, “Temporal
Entity-Relationship Models – A Survey”, IEEE
Trans. On Knowledge and Data Engineering,
11, 1999, 464 – 497.
[7] A. Gustavsson. “Maintaining the Evaluation of
Software Objects in an Integrated
Environment”, ACM – Proc. 2nd Intl. Workshop
On Software Configuration Management, 1989,
114 – 117.
[8] A. Havewala. “The Version Control Process:
How and Why it can save your project”, Dr.
Dobb’s Journal. 24, 1999, 100 – 111.
[9] C. S. Jensen and R. T. Snodgrass. “Temporal
Data Management”, IEEE Trans. on Knowledge
and Data Engineering. 11, 1999, 36 – 44.
[10] K. Torp, C. S. Jensen and R. T. Snodgrass,
“Effective Timestamping in Database”, The
Very Large Database Journal, Vol. 8, 1999, 267
– 288.
[11] B. Knight, and J. Ma. “A General Temporal
Theory”, The Computer Journal, 37, 1994, 114
– 123.
[12] B. Knight and J. Ma. “A Temporal Database
Model Supporting Relative and Absolute Time”,
The Computer Journal. 37, 1994, 588 – 597.
[13] A. Lie. “Change Oriented Versioning in a
Software Engineering Database”, ACM – Proc.
2nd Intl. Workshop on Software Configuration
Management. 1989, 56 – 65.
[14] H. Mary. “Beyond Version Control”, Software
Magazine. 16, 1996, 45 – 47.
550
Special Issue on ICIT 2009 Conference - Applied Computing

EFFECTIVE DIGITAL FORENSIC ANALYSIS

OF THE NTFS DISK IMAGE

Mamoun Alazab, Sitalakshmi Venkatraman, Paul Watters

University of Ballarat, Australia

{m.alazab, s.venkatraman, p.watters} @ballarat.edu.au

ABSTRACT
Forensic analysis of the Windows NT File System (NTFS) could provide useful
information leading towards malware detection and presentation of digital
evidence for the court of law. Since NTFS records every event of the system,
forensic tools are required to process an enormous amount of information related
to user / kernel environment, buffer overflows, trace conditions, network stack, etc.
This has led to imperfect forensic tools that are practical for implementation and
hence become popular, but are not comprehensive and effective. Many existing
techniques have failed to identify malicious code in hidden data of the NTFS disk
image. This research discusses the analysis technique we have adopted to
successfully detect maliciousness in hidden data, by investigating the NTFS boot
sector. We have conducted experimental studies with some of the existing popular
forensics tools and have identified their limitations. Further, through our proposed
three-stage forensic analysis process, our experimental investigation attempts to
unearth the vulnerabilities of NTFS disk image and the weaknesses of the current
forensic techniques.

Keywords: NTFS, forensics, disk image, data hiding.

1 INTRODUCTION predominant operating systems in use, such as

Digital forensics is the science of identifying,
extracting, analyzing and presenting the digital
evidence that has been stored in the digital electronic
storage devices to be used in a court of law [1, 2, 3].
While forensic investigation attempts to provide full
descriptions of a digital crime scene, in computer
systems, the primary goals of digital forensic
analysis are fivefold: i) to identify all the unwanted
events that have taken place, ii) to ascertain their
effect on the system, iii) to acquire the necessary
evidence to support a lawsuit, iv) to prevent future
incidents by detecting the malicious techniques used
and v) to recognize the incitement reasons and
intendance of the attacker for future predictions [2,
4]. The general component in digital forensic process
are; acquisition, preservation, and analysis [5].
Digital electronic evidence could be described as
the information and data of investigative value that
are stored by an electric device, such evidence [6].
This research focuses on the abovementioned third
goal of acquiring the necessary evidence of
intrusions that take place on a computer system. In
particular, this paper investigates the digital forensic
techniques that could be used to analyze and acquire
evidences from the most commonly used file system
on computers, namely, Windows NT File System
(NTFS).
Today, NTFS file system is the basis of
Windows 2000, Windows XP, Windows Server 2003,
Windows Server 2008, Windows Vista, Windows 7
and even in most free UNIX distributions [7, 8, 9].
Hence, malware writers try to target on NTFS as this
could result in affecting more computer users.
Another compelling reason for witnessing a strong
relationship between computer crime and the NTFS
file system is the lack of literature that unearth the
vulnerabilities of NTFS and the weaknesses of the
present digital forensic techniques [10]. This paper
attempts to fill this gap by studying the techniques
used in the analysis of the NTFS disk image. Our
objectives are i) to explore the NTFS disk image
structure and its vulnerabilities, ii) to investigate
different commonly used digital forensic techniques
such as signatures, data hiding, timestamp, etc. and
their weaknesses, and iii) finally to suggest
improvements in static analysis of NTFS disk image.
2 FORENSIC ANALYSIS PROCESS
In this section, we describe the forensic analysis
process we had adopted to achieve the above
mentioned objectives of this research work. We
conducted an empirical study using selected digital
forensic tools that are predominantly used in practice.
Several factors such as effectiveness, uniqueness and
robustness in analyzing NTFS disk image were
considered in selecting the tools / utilities required

UbiCC Journal – Volume 4 No. 3 551

Special Issue on ICIT 2009 Conference - Applied Computing
for this empirical study. Since each utility does
some specific functionality, a collection of such tools
were necessary to perform a comprehensive set of
functionalities. Hence, the following forensic
utilities / tools were adopted to conduct the
experimental investigation in this research work:
i) Disk imaging utilities such as dd [11] or
dcfldd V1.3.4-1 [12] for obtaining sector-
by-sector mirror image of the disk;
ii) Evidence collection using utilities such as
Hexedit [13], Frhed 1.4.0[14] and Strings
V2.41[15] to introspect the binary code of
the NTFS disk image;
iii) NTFS disk analysis using software tools
such as The Sleuth KIT (TSK) 3.01[16] and
Autopsy [17] and NTFSINFO v1.0 [18] to
explore and extract intruded data as well as
hidden data for performing forensic analysis.
For the experimental investigation of the
effectiveness of the above tools, we created test data
on a Pentium (R) Core (TM) 2 Due CPU, 2.19 GHz,
2.98 of RAM with Windows XP professional that
adopts the NTFS file system partition. In this pilot
empirical study, we focused on the boot sector of the
NTFS disk image. We adopted the following three
stages to perform digital forensic analysis in a
comprehensive manner:
Stage 1: Hard disk data acquisition,
Stage 2: Evidence searching and
Stage 3: Analysis of NTFS file system.
2.1 Stage 1 - Hard Disk Data Acquisition
As the first stage in forensic analysis, we used
the dcfldd developed by Nicholas Harbour and dd
utility from George Garner to acquire the NTFS disk
image from the digital electronic storage device.
This utility was selected for investigation since it
provides simple and flexible acquisition tools. The
main advantage of using these tools is that we could
extract the data in or between partitions to a separate
file for more analysis. In addition, this utility
provides built-in MD5 hashing features. Some of its
salient features allow the analyst to calculate, save,
and verify the MD5 hash values. In digital forensic
analysis, using hashing technique is important to
ensure data integrity and to identify which values of
data have been maliciously changed as well as to
explore known data objects [19].
2.2 Stage 2 - Evidence searching
The next stage involved searching for evidences
with respect to system tampering. An evidence of
intrusion could be gained by looking for some
known signatures, timestamps as well as even
searching for hidden data [20]. In this stage, we used
the Strings command by Mark Russinovich, Frhed
hexeditor tool by Rihan Kibria and WinHex
hexeditor tool by X-Ways Software Technology AG
UbiCC Journal – Volume 4 No. 3
to detect a keyword or phrase from the disk image.
2.3 Stage 3 - Analysis of NTFS File System
In the final stage of the experimental study, we
analyzed the data obtained from the NTFS disk
image that contributed towards meaningful
conclusions of the forensic investigation. We
adopted a collection of tools such as the Sleuth Kit
(TSK), Autopsy Forensic by Brian Carrier and
NTFSINFO v1.0 from Microsoft Sysinternals by
Mark Russinovich to perform different aspects of the
NTFS file system analysis.
3 FORENSIC INVESTIGATION STEPS
Many aspects must be taken into consideration
when conducting a computer forensic investigation.
There are different approaches adopted by an
investigator while examining a crime scene. From
the literature, we find five steps adopted, such as,
Policy and procedure development, Evidence
assessment, Evidence acquisition, Evidence
examination, and documenting and reporting [26]. In
our proposed approach for the digital forensic
investigation, we adopted the following nine steps as
shown in Figure 1:
Step 1: Policy and Procedure Development – In this
step, suitable tools that are needed in the digital
scene are determined as part of administrative
considerations. All aspects of policy and procedure
development are considered to determine the mission
statement, skills and knowledge, funding, personal
requirement, evidence handling and support from
management.
Step 2: Hard Disk Acquisition – This step involves
forensic duplication that could be achieved by
obtaining NTFS image of the original disk using DD
tool command. This step is for obtaining sector-by-
sector mirror image of the disk and the output of the
image file is created as Image.dd.
Step 3: Check the Data Integrity – This step ensures
the integrity of data acquired through reporting of a
hash function. We used MD5 tool to guarantee the
integrity of the original media and the resulting
image file.
Step 4: Extract MFT in the Boot Sector – In this step,
the MFT is extracted from the boot sector. We
analyzed the MFT using WinHex hexeditor tool and
checked number of sectors allocated to the NTFS file
system using NTFSINO.
Step 5: Extract $Boot file and Backup boot sector –
In this step, the $Boot file is extracted to investigate
hidden data. We analyzed the hidden data in the
$Boot metadata file system using WinHex, TSK and
Autopsy tools.
Step 6: Compare Boot sector and Backup – A
552
Special Issue on ICIT 2009 Conference - Applied Computing
comparison of the original and backup boot sectors is
performed in this step. We obtained another 2
Images from the original Image using DD tool. The
output generated resulted in two image files named,
backupbootsector.dd and bootsector.dd. We analyzed
the two image file named backupbootsector.dd and
bootsector.dd using WinHex hex-editor tool, TSK
and Autopsy tools.
Step 7: Check the Data Integrity – In this step the
integrity of data is verified again for test of
congruence. We adopted the hashing technique
using MD5 tool for the two created image files to
check the data integrity.
Step 8: Extract the ASCII and UNICODE –This step
involves extracting the
ASCII and UNICODE characters from the binary
files in the disk image. We used the Strings
command tool and keyword search for matching text
or hexadecimal values recorded on the disk. Through
keyword search, we could find even files that
contain specific words.
Step 9: Physical Presentation – In this final step, all
the findings from the forensic investigation are
documented. It involves presenting the digital
evidence through documentation and reporting
procedures.
Figure 1: Forensic investigation steps
UbiCC Journal – Volume 4 No. 3
4 BOOT SECTOR ANALYSIS OF NTFS
4.1 NTFS Disk Image
As mentioned in the previous section, the first
step to be adopted by a digital forensic investigator is
to acquire a duplicate copy of the NTFS disk image
before beginning the analysis. This is to ensure that
the data on the original devices have not been
changed during the analysis. Therefore, it is required
to isolate the original infected computer from the
disk image in order to extract the evidence that could
be found on the electronic storage devices. By
conducting investigations on the disk image, we
could unearth any hidden intrusions since the image
captures the invisible information as well [21]. The
advantages of analyzing disk images are that the
investigators can: a) preserve the digital crime-scene,
b) obtain the information in slack space, c) access
unallocated space, free space, and used space, d)
recover file fragments, hidden or deleted files and
directories, e) view the partition structure and f) get
date-stamp and ownership of files and folders [3, 22].
4.2 Master File Table
To investigate how intrusions result in data
hiding, data deletion and other obfuscations, it is
essential to understand the physical characteristics of
the Microsoft NTFS file system. Master File Table
(MFT) is the core of NTFS since it contains details
of every file and folder on the volume and allocates
two sectors for every MFT entry [23]. Hence, a good
knowledge of the MFT layout structure also
facilitates the disk recovery process. Each MFT entry
has a fixed size which is 1 KB (at byte offset 64 in
the boot sector one could identify the MFT record
size). We provide the MFT layout and represent the
plan of the NTFS file system using Figure 2. The
main purpose of NTFS is to facilitate reading and
writing of the file attributes and the MFT enables a
forensic analyst to examine in some detail the
structure and working of the NTFS volume.
Therefore, it’s important to understand how the
attributes are stored in the MFT entry.
The key feature to note is that MFT entry within
the MFT contains attributes that can have any format
and any size. Further, as shown in Figure 2, every
attribute contains an entry header which is allocated
in the first 42 bytes of a file record, and it contains an
attribute header and attribute content. The attribute
header is used to identify the size, name and the flag
value. The attribute content can reside in the MFT
followed by the attribute header if the size is less
than 700 bytes (known as a resident attribute),
otherwise it will store the attribute content in an
external cluster called cluster run (known as a non-
resident attribute). This is because; the MFT entry is
1KB in size and hence cannot fit anything that
occupies more than 700 bytes.
553
Special Issue on ICIT 2009 Conference - Applied Computing

the steps in Figure 1 to analyze the boot sector image.

As shown in Figure 3, we performed an analysis of
the data structure of this boot sector and the results
of the investigation conducted using existing forensic
tools is summarized in Table 2. From these results,
we could conclude that the existing forensic tools do
not check possible infections that could take place in
certain hidden data of the boot sector. Hence, we
describe the hidden data analysis technique that we
had adopted in the next section.

5 HIDDEN DATA ANALYSIS AND RESULTS

The recent cyber crime trends are to use different

Figure 2: MFT layout structure obfuscated techniques such as disguising file names,
hiding attributes and deleting files to intrude the
4.3 Boot Sector Analysis and Results computer system. Since the Windows operating
We performed boot sector analysis by system does not zero the slack space, it becomes a
investigating metadata files that are used to describe vehicle to hide data, especially in $Boot file. Hence,
the file system. We followed the steps described in in this study, we have analyzed the hidden data in the
previous section (Figure 1) by first creating a NTFS $Boot file structure. The $Boot entry is stored in a
disk image of the test computer using the dd utility metadata file at the first cluster in sector 0 of the file
for investigating the boot sector. We used system, called $Boot, from where the system boots.
NTFSINFO tool on the disk image as shown in Table It is the only metadata file that has a static location
1 which shows the boot sector of the test device and so that it cannot be relocated. Microsoft allocates the
information about the on-disk structure. Such data first 16 sectors of the file system to $Boot and only
structure examination enables us to view the MFT half of these sectors contains non-zero values [3].
information, allocation size, volume size and In order to investigate the NTFS file system, one
metadata files. We extracted useful information such requires to possess substantial knowledge and
as the size of clusters, sector numbers in the file experience to analyze the data structure and the
system, starting cluster address of the MFT, the size hidden data [24]. The $Boot metadata file structure is
of each MFT entry and the serial number given for located in MFT entry 7 and contains the boot sector
the file system. of the file system. It contains information about the
size of the volume, clusters and the MFT. The $Boot
Table 1: NTFS Information Details. metadata file structure has four attributes, namely,
$STANDARD_INFORMATION, $FILE_NAME,
Volume Size $SECURITY_DESCRIPTION and $DATA. The
----------- $STANDARD_INFORMATION attribute contains
Volume size : 483 MB temporal information such as flags, owner, security
Total sectors : 991199 ID and the last accessed, written, and created times.
Total clusters : 123899 The $FILE_NAME attribute contains the file name
Free clusters : 106696 in UNICODE, the size and temporal information as
Free space : 416 MB (86% of drive) well. The $SECURITY_DESCRIPTION attribute
Allocation Size contains information about the access control and
---------------- security properties. Finally, the $DATA attribute
Bytes per sector : 512 contains the file contents. These attributes values for
Bytes per cluster : 4096 the test sample are shown in Table 2 as an
Bytes per MFT record : 1024 illustration. To achieve this, we used the following
Clusters per MFT record: 0 TSK command tools:
MFT Information
--------------- Istat –f ntfs c:\image.dd 7
MFT size : 0 MB (0% of drive)
MFT start cluster : 41300 From our investigations of the resulting attribute
MFT zone clusters : 41344 - 56800 values, we find that, the $Boot data structure of the
MFT zone size : 60 MB (12% of drive) NTFS file system could be used to hide data. By
MFT mirror start : 61949 analyzing the hidden data in the boot sector, one
Meta-Data files could provide useful information for digital forensics.
The size of the data that could be hidden in the boot
sector is limited by the number of non-zero that
From the information gained above, we followed

UbiCC Journal – Volume 4 No. 3 554

Special Issue on ICIT 2009 Conference - Applied Computing
Microsoft allocated in the first 16 sectors of the file
system. The data could be hidden in the $Boot
metadata files without raising suspicion and without
affecting the functionality of the system [25].
Table 2: Results of $Boot Analysis
MFT Entry Header Values:
Entry: 7 Sequence: 7
$LogFile Sequence Number: 0
Allocated File
Links: 1
$STANDARD_INFORMATION Attribute Values:
Flags: Hidden, System
Owner ID: 0
Created: Mon Feb 09 12:09:06 2009
File Modified: Mon Feb 09 12:09:06 2009
MFT Modified: Mon Feb 09 12:09:06 2009
Accessed: Mon Feb 09 12:09:06 2009
$FILE_NAME Attribute Values:
Flags: Hidden, System
Name: $Boot
Parent MFT Entry: 5 Sequence: 5
Allocated Size: 8192 Actual Size: 8192
Created: Mon Feb 09 12:09:06 2009
File Modified: Mon Feb 09 12:09:06 2009
MFT Modified: Mon Feb 09 12:09:06 2009
Accessed: Mon Feb 09 12:09:06 2009
Attributes:
Type: $STANDARD_INFORMATION (16-0)
Name: N/A Resident size: 48
Type: $FILE_NAME (48-2) Name: N/A Resident
size: 76
Type: $SECURITY_DESCRIPTOR (80-3)
Name: N/A Resident size: 116
Type: $DATA (128-1) Name: $Data Non-
Resident size: 8192
01
Analysis of the $Boot data structure of the NTFS
file system will identify any hidden data. The
analyzer should start by making a comparison
between the boot sector and the backup boot sector.
The image with the boot sector and backup boot
sector are supposed to be identical; otherwise there is
some data hidden in the $Boot data structure. One
method is to check the integrity of the backup boot
sector and the boot sector by calculating the MD5 for
both of them. A difference in checksum indicates
that there is some hidden data. We performed this
comparison by adopting the following commands on
the $Boot image file and the backup boot image:
dd if=image.dd bs=512 count=1 skip=61949
of=c:\backupbootsector.dd –md5sum –verifymd5 –
UbiCC Journal – Volume 4 No. 3
md5out=c:\hash1.md5
dd if=image.dd bs=512 count=1 of=c:\bootsector.dd
–md5sum –verifymd5 –md5out=c:\hash2.md5
We found that hidden data in the $Boot data
structure could not be detected directly by the
existing tools used in this study and manual
inspections were required alongside these forensic
tools. Hence, through the analysis conducted with
various existing utilities and tools, we arrived at the
following results:
i) Since NTFS stores all events that take place
on a computer system, there is a huge amount
of data analysis required while scanning the
entire NTFS disk image for forensic purposes.
In this empirical study, by merely focusing
on the hidden data of the $Boot file, we have
shown that a variety of tools and utilities had
to be adopted along with manual inspections.
Hence, it takes an enormous amount of time
to analyze the data derived with such tools.
ii) The existing forensic tools are not
comprehensive and effective in identifying
the recent computer threats. Not all computer
infections are detected by forensic tools,
especially intrusions that are in the form of
hidden data in the $Boot file go unchecked.
iii) It was mandatory to perform manual
investigations alongside the existing tools. By
adopting a manual introspection of the $Boot
file using the three-stage approach of i) hard
disk acquisition, ii) evidence searching and
iii) analysis of the NTFS file system, we
could successfully identify hidden data in the
$Boot file.
iv) Intelligent search techniques could be adopted
to extract the ASCII and UNICODE
characters from binary files in the disk image
on either the full file system image or just the
unallocated space, which could speed-up the
process of identifying hidden data.
v) One of the main reasons for having varying
tools is that Microsoft has different versions
of the NTFS file system to be catered for.
While Windows XP and Windows Server
2003 use the same NTFS version, Windows
Vista uses the NTFS 3.1 version [7]. The new
NTFS 3.1 has changed the on-disk structure.
For example, the location of the volume boot
record is at physical sector 2,048. Most of the
existing tools do not work with all the
different versions of NTFS file system, and
hence a comprehensive tool is warranted to
cope with these changes.
555
Special Issue on ICIT 2009 Conference - Applied Computing

Figure 3: Analysis of the test boot Sector

Table 2: Results from the analysis of the test boot sector.

Byte
Size Description Value Action / Result
Range
If bootable, jump. If non-bootable,
0 -- 2 3 Jump to boot code 9458411
used to store error message
3 -- 10 8 OEM Name – System ID NTFS
11 -- 12 2 Bytes per sector: 512
13 -- 13 1 Sectors per cluster 8
14 -- 15 2 Reserved sectors 0 Unused – Possible Infection
16 -- 20 5 Unused 0 Unused – Possible Infection
21 -- 21 1 Media descriptor 0
22 -- 23 2 Unused 0 Unused – Possible Infection
24 -- 25 2 Sectors per track 63 No Check – Possible Infection
26 -- 27 2 Number of heads 255 No Check – Possible Infection
28 -- 31 4 Unused 32 No Check – Possible Infection
32 -- 35 4 Unused 0 Unused – Possible Infection
36 -- 39 4 Drive type check 80 00 00 00 For USB thumb drive
Number of sectors in file
40 -- 47 8 0.47264 GB
system (volume)
Starting cluster address of
48 -- 55 8 4*8=32
$MFT
Starting cluster address of MFT
56 -- 63 8 619,49
Mirror $DATA attribute
64 -- 64 1 Size of record - MFT entry 210=1024
65 -- 67 3 Unused 0 Unused – Possible Infection
68 -- 68 1 Size of index record 01h
69 -- 71 3 Unused 0 Unused – Possible Infection
72 -- 79 8 Serial number C87C8h
80 -- 83 4 Unused 0 Unused – Possible Infection
84 -- 509 426 Boot code ~
510 --511 2 Boot signature 0xAA55

UbiCC Journal – Volume 4 No. 3 556

Special Issue on ICIT 2009 Conference - Applied Computing

6 CONCLUSIONS AND FUTURE Digital forensic techniques for static analysis

RESEARCH DIRECTIONS of NTFS images, Proceedings of ICIT2009,
Fourth International Conference on
Recent methods adopted by computer intruders, Information Technology, IEEE Xplore (2009).
attackers and malwares are to target hidden and [3] B. Carrier: File system forensic analysis,
deleted data so that they could evade from virus Addison-Wesley Professional, USA, (2008).
scanners and become even difficult to be identified [4] S. Ardisson: Producing a Forensic Image of
using existing digital forensic tools. This paper has Your Client’s Hard Drive? What You Need to
attempted to explore the difficulties involved in Know, Qubit, 1, pp. 1-2 (2007).
digital forensics, especially in conducting NTFS [5] M. Andrew: Defining a Process Model for
disk image analysis and to propose an effective Forensic Analysis of Digital Devices and
digital forensic analysis. Storage Media, Proceedings of SADFE2007,
In this empirical study, we have found that the Second International Workshop on Systematic
boot sector of the NTFS file system could be used Approaches to Digital Forensic Engineering,
as a vehicle to hide data by computer attackers as pp. 16-30 (2007).
there is a potential weakness. We have emphasized [6] E Investigation: Electronic Crime Scene
the knowledge and importance of file systems for Investigation: A Guide for First Responders,
digital forensics, as several techniques to hide data US Department of Justice, NCJ, (2001).
such as slack space and hidden attributes are being [7] Svensson, A., “Computer Forensic Applied to
recently adopted by attackers. This is an important Windows NTFS Computers”, Stockholm's
NTFS file system weakness to be addressed and University, Royal Institute of Technology,
research in this domain area could lead to effective (2005).
solution for the open problem of detecting new [8] NTFS, http://www.ntfs.com, 22/2/2009.
malicious codes that make use of such an [9] D. Purcell & S. Lang: Forensic Artifacts of
obfuscated mode of attack. We have shown that Microsoft Windows Vista System, Lecture
the existing forensic software tools are not Notes in Computer Science, Springer, 5075,
competent enough to comprehensively detect all pp. 304-319 (2008).
hidden data in boot sectors. [10] T. Newsham, C. Palmer, A; Stamos & J.
As a first step to address this problem, we have Burns: Breaking forensics software:
proposed a three-stage forensic analysis process Weaknesses in critical evidence collection,
consisting of nine steps to facilitate the Proceedings of the 2007 Black Hat
experimental study. We have reported the results Conference, (2007).
gathered by following these proposed steps. By [11] DD tool, George Garner’s site, Retrieved
adopting effective search techniques, we were January, 2009 from
successful in identifying some unknown malicious http://users.erols.com/gmgarner/forensics/.
hidden data in the $Boot file that were undetected [12] DCFL tool, Nicholas Harbour,
by current forensic tools. http://dcfldd.sourceforge.net/, accessed on
In this pilot study we had adopted a few 14/1/2009.
forensic techniques and effective manual [13] WinHex tool, X-Ways Software Technology
inspections of the NTFS file image. Our future AG, Retrieved January, 2009 from
research directions would be to automate the http://www.x-ways.net/winhex/.
proposed process so as to facilitate forensic analysis [14] FRHED tool, Raihan Kibria site,
of the NTFS disk image in an efficient and http://frhed.sourceforge.net/, 14/1/2009.
comprehensive manner. We plan to extract and [15] STRINGS, Mark Russinovich, Retrieved
extrapolate malware signatures effectively as well January, 2009 from
as intelligently for any existing and even new http://technet.microsoft.com/en-
malware that use hidden and obfuscated modes of us/sysinternals/bb897439.aspx.
attack. We would automate the knowledge of how [16] TSK tools, Brian Carrier site,
to extract data from hidden data structures and how http://www.sleuthkit.org/sleuthkit/, 14/1/2009.
to reclaim deleted data and we believe this would [17] Autopsy tools, Brian Carrier site, Retrieved
extensively benefit the digital evidence collection January, 2009 from
and recovery process. http://www.sleuthkit.org/autopsy/.
[18] NTFSINFO tool, Mark Russinovich,
7 REFERENCES Retrieved January, 2009 from
http://technet.microsoft.com/en-
[1] M. Reith, C. Carr, & G. Gunsch: An au/sysinternals/bb897424.aspx.
examination of digital forensic models, [19] V. Roussev, Y.Chen, T. Bourg & G. Richard:
International Journal of Digital Evidence, 1, Forensic file system hashing revisited, Digital
pp. 1-12 (2002). Investigation, Elsevier, 3, pp. 82-90 (2006).
[2] M. Alazab, S. Venkatraman & P. Watters: [20] K. Chow, F. Law, M. Kwan & K. Lai: The

UbiCC Journal – Volume 4 No. 3 557

Special Issue on ICIT 2009 Conference - Applied Computing

Rules of Time on NTFS File System,

Proceedings of the Second International
Workshop on Systematic Approaches to
Digital Forensic Engineering, pp. 71-85(2007).

[21] K.; Jones, R. Bejtlich & C. Rose: Real digital

forensics: computer security and incident
response, Addison-Wesley Professional, USA,
(2008).
[22] H. Carvey: Windows Forensic Analysis DVD
Toolkit, Syngress Press, USA, (2007).
[23] L. Naiqi, W. Yujie & H. QinKe: Computer
Forensics Research and Implementation Based
on NTFS File System, CCCM'08, ISECS
International Colloquium on Computing,
Communication, Control, and Management,
(2008).
[24] J. Aquilina, E. Casey & C. Malin: Malware
Forensics Investigating and Analyzing
Malicious Code, Syngress Publishing,USA,
(2008).
[25] E. Huebner, D. Bem & C., Wee: Data hiding
in the NTFS file system”, Digital
Investigation, Elsevier, (2006), 3, 211-226.
[26] S. Hart, J. Ashcroft & D. Daniels:
Forensic examination of digital evidence: a
guide for law enforcement,
National Institute of Justice NIJ-US,
Washington DC, USA, Tech. Rep. NCJ,
(2004).

UbiCC Journal – Volume 4 No. 3 558

Special Issue on ICIT 2009 Conference - Applied Computing

JOB AND APPLICATION-LEVEL SCHEDULING IN DISTRIBUTED

COMPUTING

Victor V. Toporkov
Computer Science Department, Moscow Power Engineering Institute,
ul. Krasnokazarmennaya 14, Moscow, 111250 Russia
ToporkovVV@mpei.ru

ABSTRACT
This paper presents an integrated approach for scheduling in distributed computing
with strategies as sets of job supporting schedules generated by a critical works
method. The strategies are implemented using a combination of job-flow and
application-level techniques of scheduling within virtual organizations of Grid.
Applications are regarded as compound jobs with a complex structure containing
several tasks co-allocated to processor nodes. The choice of the specific schedule
depends on the load level of the resource dynamics and is formed as a resource
request, which is sent to a local batch-job management system. We propose
scheduling framework and compare diverse types of scheduling strategies using
simulation studies.

Keywords: distributed computing, scheduling, application level, job flow,

metascheduler, strategy, supporting schedules, task, critical work.

1 INTRODUCTION allows adapting resource usage and optimizing a

The fact that a distributed computational
environment is heterogeneous and dynamic along
with the autonomy of processor nodes makes it much
more difficult to manage and assign resources for job
execution at the required quality level [1].
When constructing a computing environment
based on the available resources, e.g. in the model
which is used in X-Com system [2], one normally
does not create a set of rules for resource allocation
as opposed to constructing clusters or Grid-based
virtual organizations. This reminds of some
techniques, implemented in Condor project [3, 4].
Non-clustered Grid resource computing
environments are using similar approach. For
example, @Home projects which are based on
BOINC system realize cycle stealing, i.e. either idle
computers or idle cycles of a specific computer.
Another still similar approach is related to the
management of distributed computing based on
resource broker assignment [5-11]. Besides Condor
project [3, 4], one can also mention several
application-level scheduling projects: AppLeS [6],
APST [7], Legion [8], DRM [9], Condor-G [10], and
Nimrod/G [11].
It is known, that scheduling jobs with
independent brokers, or application-level scheduling,
schedule for the specific job, for example, decreasing
its completion time. Such approaches are important,
because they take into account details of job structure
and users resource load preferences [5]. However,
when independent users apply totally different
criteria for application optimization along with job-
flow competition, it can degrade resource usage and
integral performance, e.g. system throughput,
processor nodes load balance, and job completion
time.
Alternative way of scheduling in distributed
computing based on virtual organizations includes a
set of specific rules for resource use and assignment
that regulates mutual relations between users and
resource owners [1]. In this case only job-flow level
scheduling and allocation efficiency can be
increased. Grid-dispatchers [12] or metaschedulers
are acting as managing centres like in the GrADS
project [13]. However, joint computing nature of
virtual organizations creates a number of serious
challenges. Under such conditions, when different
applications are not isolated, it is difficult to achieve
desirable resource performance: execution of the
user’s processes can cause unpredictable impact on
other neighbouring processes execution time.
Therefore, there are researches that pay attention to
the creation of virtual machine based virtual Grid

UbiCC Journal – Volume 4 No. 3 559

Special Issue on ICIT 2009 Conference - Applied Computing
workspaces by means of specialized operating
systems, e.g., in the new European project XtreemOS
(http://www.xtreemos.eu).
Inseparability of the resources makes it much
more complicated to manage jobs in a virtual
organization, because the presence of local job-flows
launched by owners of processor nodes should be
taken into account. Dynamical load balance of
different job-flows can be based on economical
principles [14] that support fairshare division model
for users and owners. Actual job-flows presence
requires forecasting resource state and their
reservation [15], for example by means of Maui
cluster scheduler simulation approach or methods,
implemented in systems such as GARA, Ursala, and
Silver [16].
The above-mentioned works are related to either
job-flow scheduling problems or application-level
scheduling.
Fundamental difference between them and the
approach described is that the resultant dispatching
strategies are based on the integration of job-flows
management methods and compound job scheduling
methods on processor nodes. It allows increasing the
quality of service for the jobs and distributed
environment resource usage efficiency.
It is considered, that the job can be compound
(multiprocessor) and the tasks, included in the job,
are heterogeneous in terms of computation volume
and resource need. In order to complete the job, one
should co-allocate the tasks to different nodes. Each
task is executed on a single node and it is supposed,
that the local management system interprets it as a
job accompanied by a resource request.
On one hand, the structure of the job is usually
not taken into account. The rare exception is the
Maui cluster scheduler [16], which allows for a
single job to contain several parallel, but
homogeneous (in terms of resource requirements)
tasks. On the other hand, there are several resource-
query languages. Thus, JDL from WLMS
(http://edms.cern.ch) defines alternatives and
preferences when making resource query, ClassAds
extensions in Condor-G [10] allows forming
resource-queries for dependant jobs. The execution
of compound jobs is also supported by WLMS
scheduling system of gLite platform
(http://www.glite.org), though the resource
requirements of specific components are not taken
into account.
What sets our work apart from other scheduling
research is that we consider coordinated application-
level and job-flow management as a fundamental part
of the effective scheduling strategy within the virtual
organization.
Environment state of distribution, dynamics of its
configuration, user’s and owner’s preferences cause
the need of building multifactor and multicriteria job
UbiCC Journal – Volume 4 No. 3
managing strategies [17-20]. Availability of
heterogeneous resources, data replication policies
[12, 21, 22] and multiprocessor job structure for
efficient co-allocation between several processor
nodes should be taken into account.
In this work, the multicriteria strategy is regarded
as a set of supporting schedules in order to cover
possible events related to resource availability.
The outline of the paper is as follows.
In section 2, we provide details of application-
level and job-flow scheduling with a critical works
method and strategies as sets of possible supporting
schedules.
Section 3 presents a framework for integrated
job-flow and application-level scheduling.
Simulation studies of coordinated scheduling
techniques and results are discussed in Section 4.
We conclude and point to future directions in
Section 5.
2 APPLICATION-LEVEL AND JOB-FLOW
SCHEDULING STRATEGIES
2.1 Application-Level Scheduling Strategy
The application-level scheduling strategy is a set
of possible resource allocation and supporting
schedules (distributions) for all N tasks in the job
[18]:
Distribution:=
< <Task 1/Allocation i,
[Start 1, End 1]>, …,
<Task N/Allocation j,
[Start N, End N]> >,
where Allocation i, j is the processor node i,
j for Task 1, N; Start 1, N, End 1, N – run
time and stop time for Task 1, N execution.
Time interval [Start, End] is treated as so
called walltime (WT), defined at the resource
reservation time [15] in the local batch-job
management system.
Figure 1 shows some examples of job graphs in
strategies with different degrees of distribution, task
details, and data replication policies [19]. The first
type strategy S1 allows scheduling with fine-grain
computations and multiple data replicas, the second
type strategy S2 is one with fine-grain computations
and a bounded number of data replicas, and the third
type S3 implies coarse-grain computations and
constrained data replication. The vertices P1, …, P6,
P23, and P45 correspond to tasks, while D1, …, D8,
D12, D36, and D78 correspond to data
transmissions. The transition from graph G1 to
graphs G2 and G3 is performed through lumping of
tasks and reducing of the parallelism level.
The job graph is parameterized by prior estimates
of the duration Tij of execution of a task Pi for a
560
Special Issue on ICIT 2009 Conference - Applied Computing

processor node nj of the type j, of relative volumes The processor node load level LLj is the ratio of
Vij of computations on a processor (CPU) of the the total time of usage of the node of the type j to
type j, etc. (Table 1). the job run time. Schedules in Fig. 2, b and Fig. 2, c
It is to mention, such estimations are also are related to strategies S2 and S3.
necessary in several methods of priority scheduling
including backfilling in Maui cluster scheduler. 2.2 Critical Works Method
Strategies are generated with a critical works
P2 D3 P4 G1 method [20].
D1
D4
D7 The gist of the method is a multiphase procedure.
The first step of any phase is scheduling of a critical
P1 D5 P6 work – the longest (in terms of estimated execution
P3 P5
D2 D8
time Tij for task Pi) chain of unassigned tasks
D6 along with the best combination of available
G2
P2 P4 resources. The second step is resolving collisions
cased by conflicts between tasks of different critical
works competing for the same resource.
P1 P6 (a)
D12 D36 D78 Nodes
n1 P1 P2 P4 LL1=0.35
P3 P5 G3 n2 P5 LL2=0.10
n3 P3 LL3=0.15
n4 P6 LL4=0.50
D12 D36 D78
Nodes CF=41
P1 P6
P23 P45 n1 P1 P2 P6 LL1=0.35
n2
Figure 1: Examples of job graphs. n3 P3 P4
LL2=0
LL3=0.65
n4 P5 LL4=0.50
Figure 2 shows fragments of strategies of types Nodes CF=37

S1, S2, and S3 for jobs in Fig. 1. n1 P2 P4 P6 LL1=0.35

n2 P5 LL2=0.10
The duration of all data transmissions is equal to n3 P3 LL3=0.15
one unit of time for G1, while the transmissions D12 n4 P1 LL4=0.50
CF=41
and D78 require two units of time and the
20 Time
transmission D36 requires four units of time for G2 0 5 10 15
(b)
and G3. Nodes
We assume that the lumping of tasks is n1 P1 P2 P4 P6 LL1=0.60
characterized by summing of the values of n2 LL2=0
n3 P3 LL3=0.15
corresponding parameters of constituent subtasks n4 P5 LL4=0.20
(see Table 1). CF=39
0 5 10 15 20 Time
Table 1: User's task estimations. (c)
Nodes
Tij, Tasks n1 P1 P23 P45 P6 LL1=1
Vij P1 P2 P3 P4 P5 P6 n2 LL2=0
Ti1 2 3 1 2 1 2 n3
n4
LL3=0
LL4=0
Ti2 4 6 2 4 2 4 CF=25
Ti3 6 9 3 6 3 6 0 5 10 15 20 Time
Ti4 8 12 4 8 4 8
Vij 20 30 10 20 10 20 Figure 2: Fragments of scheduling strategies S1 (a),
S2 (b), S3 (c).
Supporting schedules in Fig. 2, a present a subset
of a Pareto-optimal strategy of the type S1 for tasks For example, there are four critical works 12, 11,
Pi, i=1, …, 6 in G1. 10, and 9 time units long (including data transfer
The Pareto relation is generated by a vector of time) on fastest processor nodes of the type 1 for the
criteria CF, LLj, j=1, …, 4. job graph G1 in Fig. 1, a (see Table 1):
A job execution cost-function CF is equal to the
sum of Vij/Ti, where Ti is the real load time of (P1, P2, P4, P6), (P1, P2, P5, P6),
processor node j by task Pi rounded to nearest not-
smaller integer. Obviously, actual solving time Ti (P1, P3, P4, P6), (P1, P3, P5, P6).
for a task can be different from user estimation Tij The schedule with CF=37 has a collision (see Fig.
(see Table 1). 2, a), which occurred due to simultaneous attempts of

UbiCC Journal – Volume 4 No. 3 561

Special Issue on ICIT 2009 Conference - Applied Computing

tasks P4 and P5 to occupy processor node n4. This
collision is further resolved by the allocation of P4
to the processor node n3 and P5 to the node n4.
Such reallocations can be based on virtual
organization economics – in order to take higher
performance processor node, user should “pay”
more. Cost-functions can be used in economical
models [14] of resource distribution in virtual
organizations. It is worth noting that full costing in
CF is not calculated in real money, but in some
conventional units (quotas), for example like in
corporate non-commercial virtual organizations. The
essential point is different – user should pay
additional cost in order to use more powerful
resource or to start the task faster. The choice of a
specific schedule from the strategy depends on the
state and load level of processor nodes, and data
storage policies.
2.3 Examples of Scheduling Strategies
Let us assume that we need to construct a
conditionally optimal strategy of the distribution of
processors according to the main scheme of the
critical works method from [20] for a job represented
by the information graph G1 (see Fig. 1). Prior
estimates for the duration Tij of processing tasks
P1, …, P6 and relative computing volumes Vij for
four types of processors are shown in Table 1, where
i = 1, …, 6; j = 1, …, 4. The number of processors
of each type is equal to 1. The duration of all data
exchanges D1, …, D8 is equal to one unit of time.
The walltime is given to be WT = 20. The criterion of
resource-use efficiency is a cost function CF. We
take a prior estimate for the duration Tij that is the
nearest to the limit time Ti for the execution of task
Pi on a processor of type j, which determines the
type j of the processor used.
The conflicts between competing tasks are
resolved through unused processors, which, being
used as resources, are accompanied with a minimum
value of the penalty cost function that is equal to the
sum of Vij/Tij (see Table 1) for competing tasks.
It is required to construct a strategy that is
conditionally minimal in terms of the cost function
CF for the upper and lower boundaries of the
maximum range for the duration Tij of the
execution of each task Pi (see Table 1). It is a
modification of the strategy S1 with fine-grain
computations, active data replication policy, and the
best- and worst execution time estimations.
The strategy with a conditional minimum with
respect to CF is shown in Table 2 by schedules 1, 2,
and 3 (Ai is allocation of task Pi, i = 1, …, 6) and
the scheduling diagrams are demonstrated in Fig. 2,
a.
The strategies that are conditionally maximal with
respect to criteria LL1, LL2, LL3, and LL4 are given
in Table 2 by the cases 4-7; 8, 9; 10, 11; and 12-14,
respectively. Since there are no conditional branches
in the job graph (see Fig. 1), LLj is the ratio of the
total time of usage of a processor of type j to the
walltime WT of the job completion.
The Pareto-optimal strategy involves all
schedules in Table 2. The schedules 2, 5, and 13
have resolved collisions between tasks P4 and P5.
Let us assume that the load of processors is such
that the tasks P1, P2, and P3 can be assigned with
no more than three units of time on the first and third
processors (see Table 2). The metascheduler runs
through the set of supporting schedules and chooses a
concrete variant of resource distribution that depends
on the actual load of processor nodes.

Table 2: The strategy of the type MS1.

Sche- Duration Allocation Criteria
dule
T1 T2 T3 T4 T5 T6 A1 A2 A3 A4 A5 A6 CF LL1 LL2 LL3 LL4
1 2 3 3 2 2 10 1 1 3 1 2 4 41 0.35 0.10 0.15 0.50
2 2 3 3 10 10 2 1 1 3 3 4 1 37 0.35 0 0.65 0.50
3 10 3 3 2 2 2 4 1 3 1 2 1 41 0.35 0.10 0.15 0.50
4 2 3 3 2 2 10 1 1 3 1 2 1 41 0.85 0.10 0.15 0
5 2 3 3 10 10 2 1 1 3 4 1 1 38 0.85 0 0.15 0.50
6 2 11 11 2 2 2 1 4 1 1 2 1 39 0.85 0.10 0 0.55
7 10 3 3 2 2 2 1 1 3 1 2 1 41 0.85 0.10 0.15 0
8 2 11 11 2 2 2 1 4 2 1 2 1 39 0.30 0.65 0 0.55
9 10 3 3 2 2 2 2 1 2 1 2 1 41 0.35 0.75 0 0
10 2 11 11 2 2 2 1 3 4 1 2 1 41 0.30 0.10 0.55 0.55
11 10 3 3 2 2 2 3 1 3 1 2 1 41 0.35 0.10 0.60 0
12 2 3 3 2 2 10 1 1 3 1 2 4 41 0.35 0.10 0.15 0.50
13 2 3 3 10 10 2 1 1 3 3 4 1 39 0.35 0 0.65 0.50
14 10 3 3 2 2 2 4 1 3 1 2 1 41 0.35 0.10 0.15 0.50

UbiCC Journal – Volume 4 No. 3 562

Special Issue on ICIT 2009 Conference - Applied Computing

Then, the metascheduler should choose the
schedules 1, 2, 4, 5, 12, and 13 as possible variants
of resource distribution. However, the concrete
schedule should be formulated as a resource request
and implemented by the system of batch processing
subject to the state of all four processors and possible
runtimes of tasks P4, P5, and P6 (see Table 2).
Suppose that we need to generate a Pareto-
optimal strategy for the job graph G2 (see Fig. 1) in
the whole range of the duration Ti of each task Pi,
while the step of change is taken to be no less than
the lower boundary of the range for the most
performance processor.
The Pareto relation is generated by the vector of
criteria CF, LL1, … , LL4. The remaining initial
conditions are the same as in the previous example.
The strategies that are conditionally optimal with
respect to the criteria CF, LL1, LL2, LL3, and LL4
are presented in Table 3 by the schedules 1-4, 5-12,
13-17, 18-25, and 26-33, respectively. The Pareto-
optimal strategy does not include the schedules 2, 5,
12, 14, 16, 17, 22, and 30.
Let us consider the generation of a strategy for
the job represented structurally by the graph G3 in
Fig. 1 and by summing of the values of the
parameters given in Table 1 for tasks P2, P3 and P4,
P5.
As a result of the resource distribution for the
model G3, the tasks P1, P23, P45, and P6 turn out
to be assigned to one and the same processor of the
first type. Consequently, the costs of data exchanges
D12, D36, and D78 can be excluded. Because there
can be no conflicts in this case between processing
tasks (see Fig. 1), the scheduling obtained before the
exclusion of exchange procedures can be revised.

Table 3: The strategy of the type S2.

Sche- Duration Allocation Criteria
dule
T1 T2 T3 T4 T5 T6 A1 A2 A3 A4 A5 A6 CF LL1 LL2 LL3 LL4
1 2 3 3 4 4 3 1 1 3 2 4 1 39 0.40 0.20 0.15 0.20
2 4 3 3 2 2 3 2 1 3 1 2 1 41 0.40 0.30 0.15 0
3 4 3 3 3 3 2 2 1 3 1 3 1 40 0.40 0.20 0.30 0
4 5 3 3 2 2 2 2 1 3 1 2 1 43 0.35 0.35 0.15 0
5 2 3 3 2 2 5 1 1 3 1 2 1 43 0.60 0.10 0.15 0
6 2 3 3 4 4 3 1 1 3 1 4 1 39 0.60 0 0.15 0.20
7 2 3 3 5 5 2 1 1 3 1 4 1 41 0.60 0 0.15 0.25
8 2 6 6 2 2 2 1 2 1 1 2 1 42 0.60 0.30 0 0
9 4 3 3 2 2 3 1 1 3 1 2 1 41 0.60 0.10 0.15 0
10 4 3 3 3 3 2 1 1 3 1 3 1 40 0.60 0 0.30 0
11 4 4 4 2 2 2 1 1 4 1 2 1 41 0.60 0.10 0 0.20
12 5 3 3 2 2 2 1 1 3 1 2 1 43 0.60 0.10 0.15 0
13 2 6 6 2 2 2 1 2 4 1 2 1 43 0.30 0.40 0 0.30
14 4 3 3 2 2 3 2 1 2 1 2 1 41 0.40 0.45 0 0
15 4 3 3 3 3 2 2 1 2 1 2 1 40 0.40 0.50 0 0
16 4 4 4 2 2 2 2 1 2 1 2 1 41 0.40 0.50 0 0
17 5 3 3 2 2 2 2 1 2 1 2 1 43 0.35 0.50 0 0
18 2 3 3 2 2 5 1 1 3 1 2 2 43 0.35 0.35 0.15 0
19 2 3 3 4 4 3 1 1 3 2 3 1 39 0.40 0.20 0.35 0
20 2 3 3 5 5 2 1 1 3 2 3 1 40 0.35 0.25 0.40 0
21 2 6 6 2 2 2 1 2 3 1 2 1 42 0.30 0.40 0.30 0
22 4 3 3 2 2 3 2 1 3 1 2 1 41 0.40 0.30 0.15 0
23 4 3 3 3 3 2 2 1 3 1 3 1 40 0.40 0.20 0.30 0
24 4 4 4 2 2 2 2 1 3 1 2 1 41 0.40 0.30 0.20 0
25 5 3 3 2 2 2 2 1 3 1 2 1 43 0.35 0.35 0.15 0
26 2 3 3 2 2 5 1 1 3 1 2 2 43 0.35 0.35 0.15 0
27 2 3 3 4 4 3 1 1 3 2 4 1 39 0.40 0.20 0.15 0.20
28 2 3 3 5 5 2 1 1 3 2 4 1 40 0.35 0.25 0.15 0.25
29 2 6 6 2 2 2 1 2 4 1 2 1 42 0.30 0.40 0 0.30
30 4 3 3 2 2 3 2 1 3 1 2 1 41 0.40 0.30 0.15 0
31 4 3 3 3 3 2 2 1 3 1 3 1 40 0.40 0.20 0.30 0
32 4 4 4 2 2 2 2 1 4 1 2 1 41 0.40 0.30 0 0.20
33 5 3 3 2 2 2 2 1 3 1 2 1 43 0.35 0.35 0.15 0

UbiCC Journal – Volume 4 No. 3 563

Special Issue on ICIT 2009 Conference - Applied Computing

Table 4: The strategy of the type S3.

Sche- Duration Allocation Criteria
dule
T1 T23 T45 T6 A1 A23 A45 A6 CF LL1 LL2 LL3 LL4
1 2 8 6 4 1 1 1 1 25 1 0 0 0
2 4 8 3 5 1 1 1 1 24 1 0 0 0
3 6 4 6 4 1 1 1 1 24 1 0 0 0
4 8 4 3 5 1 1 1 1 27 1 0 0 0
5 10 4 3 3 1 1 1 1 29 1 0 0 0
6 11 4 3 2 1 1 1 1 32 1 0 0 0

The results of distribution of processors are
presented in Table 4 (A23, A45 are allocations,
and T23, T45 are run times for tasks P23 and
P45). Schedules 1-6 in Table 4 correspond to the
strategy that is conditionally minimal with respect
to CF with LL1 = 1. Consequently, there is no sense
in generating conditionally maximal schedules with
respect to criteria LL1, …, LL4.
schedule it runs the developed mechanisms that
optimize the whole job-flow (two jobs in this
example). In that case the metascheduler will still
try to find an optimal schedule for each single job
as described above and, at the same time, it will try
to find the most optimal job assignment so that the
average load of CPUs will be maximized on a job-
flow scale.

2.4 Coordinated Scheduling with the Critical

Works Method
The critical works method was developed for
application-level scheduling [19, 20]. However, it
can be further refined to build multifactor and
multicriteria strategies for job-flow distribution in
virtual organizations. This method is based on
dynamic programming and therefore uses some
integral characteristics, for example total resource
usage cost for the tasks that compose the job.
However the method of critical works can be
referred to the priority scheduling class. There is no
conflict between these two facts, because the
method is dedicated for task co-allocation of
compound jobs.
Let us consider a simple example. Fig. 3
represents two jobs with walltimes WT1 = 110
and WT2 = 140 that are submitted to the
distributed environment with 8 CPUs. If the jobs
are submitted one-by-one the metascheduler
(Section 3) will also schedule them one-by-one and
will guarantee that every job will be scheduled
within the defined time interval and in most
efficient way in terms of a selected cost function Figure 3: Sample jobs.
and maximize average load balance of CPUs on a
single job scale (Fig. 4). Job-flow execution will be Fig. 5 shows, that both jobs are executed within
finished at WT3 = 250. This is an example of WT4 = WT2 = 140, every data dependency is taken
application-level scheduling and no integral job- into account (e.g. for the second job: task P2 is
flow characteristics are optimized in this case. executed only after tasks P0, P4, and P1 are
To combine application-level scheduling and ready), the final schedule is chosen from the
job-flow scheduling and to fully exploit the generated strategy with the lowest cost function.
advantages of the approach proposed, one can Priority scheduling based on queues is not an
submit both jobs simultaneously or store them in efficient way of multiprocessor jobs co-allocating,
buffer and execute the scheduling for all jobs in the in our opinion. Besides, there are several well-
buffer after a certain amount of time (buffer time). known side effects of this approach in the cluster
If the metascheduler gets more than one job to systems such as LL, NQE, LSF, PBS and others.

UbiCC Journal – Volume 4 No. 3 564

Special Issue on ICIT 2009 Conference - Applied Computing
For example, traditional First-Come-First-Serve
(FCFS) strategy leads to idle standing of the
resources. Another strategy, which involves job
ranking according to the specific properties, such as
computational complexity, for example Least-
Work-First (LWF), leads to a severe resource
fragmentation and often makes it impossible to
execute some jobs due to the absence of free
resources. In distributed environments these effects
can lead to unpredictable job execution time and
thereby to unsatisfactory quality of service.
Figure 4: Consequential scheduling.
In order to avoid it many projects have
components that make schedules, which are
supported by preliminary resource reservation
mechanisms [15, 16].
Figure 5: Job-flow scheduling.
UbiCC Journal – Volume 4 No. 3
One to mention is Maui cluster scheduler, where
backfilling algorithm is implemented. Remote Grid
resource reservation mechanism is also supported in
GARA, Ursala and Silver projects [16]. Here, only
one variant of the final schedule is built and it can
become irrelevant because of changes in the local
job-queue, transporting delays etc. The strategy is
some kind of preparation of possible activities in
distributed computing based on supporting
schedules (see Fig. 2, Tables 2, 3 and 4) and
reactions to the events connected with resource
assignment and advance reservations [15, 16]. The
more factors considered as formalized criteria are
taken into account in strategy generation, the more
complete is the strategy in the sense of coverage of
possible events [18, 19]. The choice of the
supporting schedule [20] depends on the utilization
state of processor nodes, data storage and relocation
policies specific to the environment, structure of the
jobs themselves and user estimations of completion
time and resource requirements.
It is important to mention that users can submit
jobs without information about the task execution
order as required by existing schedulers like Maui
cluster scheduler were only queues are supported.
Implemented mechanisms of our approach support
a complex structure for the job, which is
represented as a directed graph, so user should only
provide data dependencies between tasks (i.e. the
structure of the job). The metascheduler will
generate the schedules to satisfy their needs by
providing optimal plans for jobs (application-level
scheduling) and the needs for the resource owners
by optimizing the defined characteristics of the job-
flow for the distributed system (job-flow
scheduling).
3 METASCHEDULING FRAMEWORK
In order to implement the effective scheduling
and allocation to heterogeneous resources, it is very
important to group user jobs into flows according to
the strategy type selected and to coordinate job-
flow and application-level scheduling. A
hierarchical structure (Fig. 6) composed of a job-
flow metascheduler and subsidiary job managers,
which are cooperating with local batch-job
management systems, is a core part of a scheduling
framework proposed in this paper. It is assumed that
the specific supporting schedule is realized and the
actual allocation of resources is performed by the
system of batch processing of jobs. This schedule is
implemented on the basis of a user resource request
with a requirement to the types and characteristics
of resources (memory and processors) and to the
system software as well as generated, for example,
by the script of the job entry instruction qsub.
Therefore, the formation and support of scheduling
565
Special Issue on ICIT 2009 Conference - Applied Computing
strategies should be conducted by the
metascheduler, an intermediary link between the job
flow and the system of batch processing.
Job-flows
i j k principles [14] of resource allocation by using
Metascheduler
Job manager Job manager
for strategy Sk for strategies Si, Sj
Computer nodes
Job manager
for strategy Si Computer nodes
Computer node domains
Figure 6: Components of metascheduling
framework.
The advantages of hierarchically organized
resources managers are obvious, e.g., the
hierarchical job-queue-control model is used in the
GrADS metascheduler [13] and X-Com system [2].
Hierarchy of intermediate servers allows decreasing
idle time for the processor nodes, which can be
inflicted by transport delays or by unavailability of
the managing server while it is dealing with the
other processor nodes. Tree-view manager structure
in the network environment of distributed
computing allows avoiding deadlocks when
accessing resources. Another important aspect of
computing in heterogeneous environments is that
processor nodes with the similar architecture,
contents, administrating policy are grouped together
under the job manager control.
Users submit jobs to the metascheduler (see Fig.
6) which distributes job-flows between processor
node domains according to the selected scheduling
and resource co-allocation strategy Si, Sj or Sk. It
does not mean, that these flows cannot “intersect”
each other on nodes. The special reallocation
mechanism is provided. It is executed on the higher-
level manager or on the metascheduler-level. Job
managers are supporting and updating strategies
based on cooperation with local managers and
simulation approach for job execution on processor
nodes. Innovation of our approach consists in
mechanisms of dynamic job-flow environment
reallocation based on scheduling strategies. The
nature of distributed computational environments
itself demands the development of multicriteria and
multifactor strategies [17, 18] of coordinated
scheduling and resource allocation.
The dynamic configuration of the environment,
large number of resource reallocation events, user’s
UbiCC Journal – Volume 4 No. 3
and resource owner’s needs as well as virtual
organization policy of resource assignment should
be taken into account. The scheduling strategy is
formed on a basis of formalized efficiency criteria,
which sufficiently allow reflecting economical
relevant cost functions and solving the load balance
problem for heterogeneous processor nodes. The
strategy is built by using methods of dynamic
programming [20] in a way that allows optimizing
scheduling and resource allocation for a set of tasks,
comprising the compound job. In contrast to
previous works, we consider the scheduling strategy
as a set of admissible supporting schedules (see
Fig. 2, Tables 2 and 3). The choice of the specific
variant depends on the load level of the resource
dynamics and is formed as a resource query, which
is sent to a local batch-job processing system.
One of the important features of our approach is
resource state forecasting for timely updates of the
strategies. It allows implementing mechanisms of
adaptive job-flow reallocation between processor
nodes and domains, and also means that there is no
more fixed task assignment on a particular
processor node. While one part of the job can be
sent for execution, the other tasks, comprising the
job, can migrate to the other processor nodes
according to the updated co-allocation strategy. The
similar schedule correction procedure is also
supported in the GrADS project [13], where
multistage job control procedure is implemented:
making initial schedule, its correction during the job
execution, metascheduling for a set of applications.
Downside of this approach is the fact, that it is
based on the creation of a single schedule, so the
metascheduler stops working when no additional
resources are available and job-queue is then set to
waiting mode. The possibility of strategy updates
allows user, being integrated into economical
conditions of virtual organization, to affect job start
time by changing resource usage costs. In fact it
means that the job-flow dispatching strategy is
modified according to new priorities and this
provides competitive functioning and dynamic job-
flow balance in virtual organization with
inseparable resources.
4 SIMULATIONS STUDIES AND
RESULTS
4.1 Simulation System
We have implemented an original simulation
environment (Fig. 7) of the metascheduling
framework (see Fig. 6) to evaluate efficiency
indices of different scheduling and co-allocation
strategies. In contrast to well-known Grid
simulation systems such as ChicSim [12] or
OptorSim [23], our simulator MetaSim generates
566
Special Issue on ICIT 2009 Conference - Applied Computing
multicriteria strategies as a number of supporting
schedules for metascheduler reactions to the events
connected with resource assignment and advance
reservations.
Strategies for more than 12000 jobs with a fixed
completion time were studied. Every task of a job
had randomized completion time estimations,
computation volumes, data transfer times and
volumes. These parameters for various tasks had
difference which was equal to 2, ..., 3. Processor
nodes were selected in accordance to their relative
performance. For the first group of “fast” nodes the
relative performance was equal to 0.66, …, 1, for
the second and the third groups 0.33, …, 0.66 and
0.33 (“slow” nodes) respectively. A number of
nodes was conformed to a job structure, i.e. a task
parallelism degree, and was varied from 20 to 30.
4.2 Types of Strategies
We have studied the strategies of the following
types:
S1 – with fine-grain computations and
active data replication policy;
S2 – with fine-grain computations and a
remote data access;
S3 – with coarse-grain computations and
static data storage;
MS1 – with fine-grain computations,
active data replication policy, and the best- and
worst execution time estimations (a modification of
the strategy S1).
The strategy MS1 is less complete than the
strategy S1 or S2 in the sense of coverage of events
in distributed environment (see Tables 2 and 3).
However the important point is the generation of a
strategy by efficient and economic computational
procedures of the metascheduler. The type S1 has
Figure 7: Simulation environment of hierarchical scheduling framework based on strategies.
UbiCC Journal – Volume 4 No. 3
more computational expenses than MS1 especially
for simulation studies of integrated job-flow and
application-level scheduling.
Therefore, in some experiments with integrated
scheduling we compared strategies MS1, S2, and
S3.
4.3 Application-Level Scheduling Study
We have conducted the statistical research of the
critical works method for application-level
scheduling with above-mentioned types of strategies
S1, S2, S3. The main goal of the research was to
estimate a forecast possibility for making
application-level schedules with the critical works
method without taking into account independent job
flows. For 12000 randomly generated jobs there
were 38% admissible solutions for S1 strategy,
37% for S2, and 33% for S3 (Fig. 8). This result is
obvious: application-level schedules implemented
by the critical works method were constructed for
available resources non-assigned to other
independent jobs.
Along with it there is a conflict distribution for
the processor nodes that have different performance
(“fast” are 2-3 times faster, than “slow” ones): 32%
for “fast” ones, 68% for “slow” ones in S1, 56%
and 44% in S2, 74% and 26% for S3 (Fig. 9). This
may be explained as follows. The higher is the task
state of distribution in the environment with active
data transfer policy, the lower is the probability of
collision between tasks on a specific resource.
In order to implement the effective scheduling
and resource allocation policy in the virtual
organization we should coordinate application and
job-flow levels of the scheduling.
567
Special Issue on ICIT 2009 Conference - Applied Computing

S1 S1

S2
S2

S3
S3
Figure 8: Percentage of admissible application-level Figure 9: Percentage of collisions for “fast”
schedules. processor nodes in application-level scheduling.

4.4 Job-Flow and Application-Level Scheduling Average node load level, %

Study 80
For each simulation experiment such factors as
job completion “cost”, task execution time, 60
scheduling forecast errors (start time estimation),
strategy live-to-time (time interval of acceptable 40
schedules in a dynamic environment), and average
load level for strategies S1, MS1, S2, and S3 were
20
studied.
Figure 10 shows load level statistics of variable
performance processor nodes which allows 0
discovering the pattern of the specific resource usage S1 S2 S3
when using strategies S1, S2, and S3 with Relative processor nodes performance
coordinated job-flow and application-levels 0.66-1 0.33-0.66 0.33
scheduling.
The strategy S2 performs the best in the term of Figure 10: Processor node load level in strategies
load balancing for different groups of processor S1, S2, and S3.
nodes, while the strategy S1 tries to occupy “slow”
nodes, and the strategy S3 - the processors with the Factor quality analysis of S2, S3 strategies for
highest performance (see Fig. 10). the whole range of execution time estimations for the

UbiCC Journal – Volume 4 No. 3 568

Special Issue on ICIT 2009 Conference - Applied Computing
selected processor nodes as well as modification
MS1, when best- and worst-case execution time
estimations were taken, is shown in Figures 11 and
12.
Relative job Relative task
completion cost execution time
1 1
0.5 0.5
0 0
MS1 S2 S3
Сost Execution time
Figure 11: Job completion cost and task execution
time in strategies MS1, S2, and S3.
Lowest-cost strategies are the “slowest” ones like
S3 (see Fig. 11); they are most persistent in the term
of time-to-live as well (see Fig. 12).
Relative Start time deviation/
time-to-live job run time
1 1
0.5 0.5
0 0
MS1 S2 S3
Time-to-live Deviation
Figure 12: Time-to-live and start deviation time in
strategies MS1, S2, and S3.
The strategies of the type S3 try to monopolize
processor resources with the highest performance and
to minimize data exchanges. Withal, less persistent
are the “fastest”, most expensive and most accurate
strategies like S2. Less accurate strategies like MS1
(see Fig. 12) provide longer task completion time,
than more accurate ones like S2 (Fig. 11), which
include more possible events, associated with
processor node load level dynamics.
5 CONCLUSIONS AND FUTURE WORK
The related works in scheduling problems are
devoted to either job scheduling problems or
application-level scheduling. The gist of the
approach described is that the resultant dispatching
strategies are based on the integration of job-flows
and application-level techniques. It allows increasing
the quality of service for the jobs and distributed
environment resource usage efficiency.
Our results are promising, but we have bear in
mind that they are based on simplified computation
UbiCC Journal – Volume 4 No. 3
scenarios, e.g., in our experiments we use FCFS
management policy in local batch-job management
systems. Afore-cited research results of strategy
characteristics were obtained by simulation of global
job-flow in a virtual organization. Inseparability
condition for the resources requires additional
advanced research and simulation approach of local
job passing and local processor nodes load level
forecasting methods development. Different job-
queue management models and scheduling
algorithms (FCFS modifications, LWF, backfilling,
gang scheduling, etc.) can be used here. Along with it
local administering rules can be implemented.
One of the most important aspects here is that
advance reservations have impact on the quality of
service. Some of the researches (particularly the one
in Argonne National Laboratory) show, that
preliminary reservation nearly always increases
queue waiting time. Backfilling decreases this time.
With the use of FCFS strategy waiting time is shorter
than with the use of LWF. On the other hand,
estimation error for starting time forecast is bigger
with FCFS than with LWF. Backfilling that is
implemented in Maui cluster scheduler includes
advanced resource reservation mechanism and
guarantees resource allocation. It leads to the
difference increase between the desired reservation
time and actual job starting time when the local
request flow is growing. Some of the quality aspects
and job-flow load balance problem are associated
with dynamic priority changes, when virtual
organization user changes execution cost for a
specific resource.
All of these problems require further research.
ACKNOWLEDGEMENT. This work was
supported by the Russian Foundation for Basic
Research (grant no. 09-01-00095) and by the State
Analytical Program “The higher school scientific
potential development” (project no. 2.1.2/6718).
6 REFERENCES
[1] I. Foster, C. Kesselman, and S. Tuecke: The
Anatomy of the Grid: Enabling Scalable
Virtual Organizations, Int. J. of High
Performance Computing Applications, Vol. 15,
No. 3, pp. 200 – 222 (2001)
[2] V.V. Voevodin: The Solution of Large
Problems in Distributed Computational Media,
Automation and Remote Control, Vol. 68, No.
5, pp. 32 – 45 (2007)
[3] D. Thain, T. Tannenbaum, and M. Livny:
Distributed Computing in Practice: the Condor
Experience, Concurrency and Computation:
Practice and Experience, Vol. 17, No. 2-4, pp.
323 - 356 (2004)
569
Special Issue on ICIT 2009 Conference - Applied Computing
[4] A. Roy and M. Livny: Condor and Preemptive
Resume Scheduling, In: J. Nabrzyski, J.M.
Schopf, and J.Weglarz (eds.): Grid resource
management. State of the art and future trends,
Kluwer Academic Publishers, pp. 135 – 144
(2003)
[5] V.V. Krzhizhanovskaya and V. Korkhov:
Dynamic Load Balancing of Black-Box
Applications with a Resource Selection
Mechanism on Heterogeneous Resources of
Grid, In: 9th International Conference on
Parallel Computing Technologies, Springer,
Heidelberg, LNCS, Vol. 4671, pp. 245 – 260
(2007)
[6] F. Berman: High-performance Schedulers, In:
I. Foster and C. Kesselman (eds.): The Grid:
Blueprint for a New Computing Infrastructure,
Morgan Kaufmann, San Francisco, pp. 279 –
309 (1999)
[7] Y. Yang, K. Raadt, and H. Casanova:
Multiround Algorithms for Scheduling
Divisible Loads, IEEE Transactions on Parallel
and Distributed Systems, Vol. 16, No. 8, pp.
1092 – 1102 (2005)
[8] A. Natrajan, M.A. Humphrey, and A.S.
Grimshaw: Grid Resource Management in
Legion,” In: J. Nabrzyski, J.M. Schopf, and
J.Weglarz (eds.): Grid resource management.
State of the art and future trends, Kluwer
Academic Publishers, pp.145 – 160 (2003)
[9] J. Beiriger, W. Johnson, H. Bivens et al.:
Constructing the ASCI Grid, In: 9th IEEE
Symposium on High Performance Distributed
Computing, IEEE Press, New York, pp. 193 –
200 (2000)
[10] J. Frey, I. Foster, M. Livny et al.: Condor-G: a
Computation Management Agent for Multi-
institutional Grids, In: 10th International
Symposium on High-Performance Distributed
Computing, IEEE Press, New York, pp. 55 –
66 (2001)
[11] D. Abramson, J. Giddy, and L. Kotler: High
Performance Parametric Modeling with
Nimrod/G: Killer Application for the Global
Grid?, In: International Parallel and Distributed
Processing Symposium, IEEE Press, New
York, pp. 520 – 528 (2000)
[12] K. Ranganathan and I. Foster: Decoupling
Computation and Data Scheduling in
Distributed Data-intensive Applications, In:
11th IEEE International Symposium on High
Performance Distributed Computing, IEEE
Press, New York, pp. 376 – 381 (2002)
[13] H. Dail, O. Sievert, F. Berman et al.:
Scheduling in the Grid Application
Development Software project, In: J.
Nabrzyski, J.M. Schopf, and J.Weglarz (eds.):
Grid resource management. State of the art and
UbiCC Journal – Volume 4 No. 3
future trends, Kluwer Academic Publishers, pp.
73 – 98 (2003)
[14] R. Buyya, D. Abramson, J. Giddy et al.:
Economic Models for Resource Management
and Scheduling in Grid Computing, J. of
Concurrency and Computation: Practice and
Experience, Vol. 14, No. 5, pp. 1507 – 1542
(2002)
[15] K. Aida and H. Casanova: Scheduling Mixed-
parallel Applications with Advance
Reservations, In: 17th IEEE International
Symposium on High-Performance Distributed
Computing, IEEE Press, New York, pp. 65 –
74 (2008)
[16] D.B. Jackson: GRID Scheduling with
Maui/Silver, In: J. Nabrzyski, J.M. Schopf, and
J.Weglarz (eds.): Grid resource management.
State of the art and future trends, Kluwer
Academic Publishers, pp. 161 – 170 (2003)
[17] K. Kurowski, J. Nabrzyski, A. Oleksiak, and J.
Weglarz: Multicriteria Aspects of Grid
Resource Management, In: J. Nabrzyski, J.M.
Schopf, and J.Weglarz (eds.): Grid resource
management. State of the art and future trends,
Kluwer Academic Publishers, pp. 271 – 293
(2003)
[18] V. Toporkov: Multicriteria Scheduling
Strategies in Scalable Computing Systems, In:
9th International Conference on Parallel
Computing Technologies, Springer,
Heidelberg, LNCS, Vol. 4671, pp. 313 – 317
(2007)
[19] V.V. Toporkov and A.S. Tselishchev: Safety
Strategies of Scheduling and Resource Co-
allocation in Distributed Computing, In: 3rd
International Conference on Dependability of
Computer Systems, IEEE CS Press, pp. 152 –
159 (2008)
[20] V.V. Toporkov: Supporting Schedules of
Resource Co-Allocation for Distributed
Computing in Scalable Systems, Programming
and Computer Software, Vol. 34, No. 3, pp.
160 – 172 (2008)
[21] M. Tang, B.S. Lee, X. Tang, et al.: The Impact
of Data Replication on Job Scheduling
Performance in the Data Grid, Future
Generation Computing Systems, Vol. 22, No.
3, pp. 254 – 268 (2006)
[22] N.N. Dang, S.B. Lim, and C.K. Yeo:
Combination of Replication and Scheduling in
Data Grids, Int. J. of Computer Science and
Network Security, Vol. 7, No. 3, pp. 304 – 308
(2007)
[23] W.H. Bell, D. G. Cameron, L. Capozza et al.:
OptorSim – A Grid Simulator for Studying
Dynamic Data Replication Strategies, Int. J. of
High Performance Computing Applications,
Vol. 17, No. 4, pp. 403 – 416 (2003)
570
Special Issue on ICIT 2009 Conference - Applied Computing

Least and greatest fixed points of

a while semantics function
Fairouz Tchier
Mathematics department,
King Saud University
P.O.Box 22452
Riyadh 11495, Saudi Arabia
ftchier@hotmail.com
May 1, 2009

Abstract 1 Relation Algebras

The meaning of a program is given by specifying the
function (from input to output) that corresponds to
the program. The denotational semantic definition,
thus maps syntactical things into functions. A re-
lational semantics is a mapping of programs to re-
lations. We consider that the input-output seman-
tics of a program is given by a relation on its set of
states. In a nondeterministic context, this relation is
calculated by considering the worst behavior of the
program (demonic relational semantics). In this pa-
per, we concentrate on while loops. We will present
some interesting results about the fixed points of the
while semantics function; f (X) = Q ∨ P 2 X where
P < ∧ Q< = Ø, by taking P := t 2 B and Q := t∼ ,
one gets the demonic semantics we have assigned to
while loops in previous papers. We will show that
the least angelic fixed point is equal to the greatest
demonic fixed point of the semantics function.
Keywords: Angelic fixed points, demonic
fixed points, demonic functions, while
loops, relational demonic semantics.
Both homogeneous and heterogeneous relation alge-
bras are employed in computer science. In this pa-
per, we use heterogeneous relation algebras whose
definition is taken from [8, 27, 28].
(1) Definition. A relation algebra A is a structure
(B, ∨, ∧, −, ◦,^ ) over a non-empty set B of elements,
called relations. The unary operations −,^ are total
whereas the binary operations ∨, ∧, ◦ are partial. We
denote by B∨R the set of those elements Q ∈ B for
which the union R ∨ Q is defined and we require that
R ∈ B∨R for every R ∈ B. If Q ∈ B∨R , we say that
Q has the same type as R. The following conditions
are satisfied.
(a) (B∨R , ∨, ∧, −) is a Boolean algebra, with zero
element 0R and universal element 1R . The
elements of B∨R are ordered by inclusion, de-
noted by ≤.
(b) If the products P ◦ R and Q ◦ R are defined,
so is P ◦ Q^ . If the products P ◦ Q and P ◦ R
are defined, so is Q^ ◦ R. If Q ◦ R exists, so
does Q ◦ P for every P ∈ B∨R .
(c) Composition is associative: P ◦ (Q ◦ R) =
(P ◦ Q) ◦ R.

UbiCC Journal – Volume 4 No. 3 571

Special Issue on ICIT 2009 Conference - Applied Computing
(d) There are elements R id and idR associated
to every relation R ∈ B. R id behaves as a
right identity and idR as a left identity for
B∨R .
(e) The Schröder rule P ◦Q ≤ R ⇔ P ^ ◦−R ≤
−Q ⇔ −R ◦ Q^ ≤ −P holds whenever one
of the three expressions is defined.
(f) 1 ◦ R ◦ 1 = 1 iff R 6= 0 (Tarski rule).
If R^ ∈ B∨R , then R is said to be homogeneous. If
all R ∈ A have the same type, the operations are all
total and A itself is said to be homogeneous.
For simplicity, the universal, zero, and identity ele-
ments are all denoted by 1, 0, id, respectively. An-
other operation that occurs in this article is the re-
flexive transitive closure R∗ . It satisfies the well-
known laws
_
R∗ = Ri and R∗ = id ∨ R ◦ R∗ = id ∨ R∗ ◦ R,
i≥0
where R0 = id and Ri+1 = R ◦ Ri . From Definition
1, the usual rules of the calculus of relations can be
derived (see, e.g., [8, 10, 28]).
The notion of Galois connections is very important
in what follows, there are many definitions of Galois
connections [?]. We choose the following
one [2].
(2) Definition. Let (S, ≤S ) and (S 0 , ≤S 0 ) be two
preordered sets. A pair (f, g) of functions, where f :
S → S 0 and g : S 0 → S, forms a Galois connections
iff the following formula holds for all x ∈ S and y ∈
S0.
f (x) ≤S 0 y ⇔ x ≤S g(y).
The function f is called the lower adjoint and g
the upper adjoint.
2 Monotypes and Related Op-
erators
In the calculus of relations, there are two ways for
viewing sets as relations; each of them has its own
advantages. The first is via vectors: a relation x is
UbiCC Journal – Volume 4 No. 3
a vector [28] iff x = x ◦ 1. The second way is via
monotypes [2]: a relation a is a monotype iff a ≤ id.
The set of monotypes {a | a ∈ B∨R }, for a given R,
is a complete Boolean lattice. We denote by a∼ the
monotype complement of a.
The domain and codomain of a relation R can be
characterized by the vectors R ◦ 1 and R^ ◦ 1, re-
spectively [15, 28]. They can also be characterized
by the corresponding monotypes. In this paper, we
take the last approach. In what follows we formally
define these operators and give some of their prop-
erties.
(3) Definition. The domain and codomain opera-
tors of a relation R, denoted respectively by R< and
R> , are the monotypes defined by the equations
(a) R< = id ∧ R ◦ 1,
(b) R> = id ∧ 1 ◦ R.
These operators can also be characterized by Galois
connections(see [2, 2]). For each relation R and each
monotype a,
R< ≤ a ⇔ R ≤ a ◦ 1,
R> ≤ a ⇔ R ≤ 1 ◦ a.
The domain and codomain operators are linked by
the equation R> = R^ < , as is easily checked.
(4) Definition. Let R be a relation and a be a
monotype. The monotype right residual and mono-
type left residual of a by R (called factors in [5]) are
defined respectively by
(a) a/•R := ((1 ◦ a)/R)> ,
(b) R\•a := (R\(a 2 1))< .
An alternative characterization of residuals can
also be given by means of a Galois connection as
follows [1]:
b ≤ a/•R ⇔ (b 2 R)> ≤ a,
b ≤ R\•a ⇔ (R ◦ b)< ≤ a.
We have to use exhaustively the complement of
the domain of a relation R, i.e the monotype a such
that a = R< ∼ . To avoid the notation R< ∼ , we adopt
the Notation
572
Special Issue on ICIT 2009 Conference - Applied Computing
R≺ := R< ∼ .
Because we assume our relation algebra to be com-
plete, least and greatest fixed points of monotonic
functions exist. We cite [12] as a general reference
on fixed points.
Let f be a monotonic function. The fol-
lowing properties
V of fixed points Vare used below:
(a) µf = W {X|f (X) = X} = W {X|f (X) ≤ X},
(b) νf = {X|f (X) = X} = {X|X ≤ f (X)},
(c) µf ≤ νf,
(d) f (Y ) ≤ Y ⇒ µf ≤ Y,
(e) Y ≤ f (Y ) ⇒ Y ≤ νf.
In what follows, we describe notions that are useful
for the description of the set of initial states of a
program for which termination is guaranteed. These
notions are progressive finiteness and the initial part
of a relation.
A relation R is progressively finite in terms of
points iff there are no infinite chains s0 , ..., si such
that si Rsi+1 ∀i, i ≥ 0. I.e there is no points set y
which are the starting points of some path of infinite
length. For every point set y, y ≤ R ◦ y ⇒ y = 0.
The least set of points which are the starting points
of paths of finite length i.e from which we can pro-
ceed only finitely many steps is called initial part
of R denoted by I(R). This topic is of interest in
many areas of computer science, mathematics and is
related to recursion and induction principle.
(5) Definition.
(a) The initial part of a relation R, denoted
I(R), is V
given by V ( 2 ). We give the definitions and needed properties
I(R) = {a | a ≤ id : a/•R = a} = {a |
a ≤ id : a/•R ≤ a} = µ(a : a ≤ id : a/•R),
where a is a monotype.
(b) A relation R is said to be progressively finite
[28] iff I(R) = id.
The description of I(R) by the formulation a/•R = a
shows that I(R) exists, since (a | a ≤ id : a/•R) is
monotonic in the first argument and because the set
of monotypes is a complete lattice, it follows from the
fixed point theorem of Knaster and Tarski that this
function has a least fixed point. Progressive finite-
ness of a relation R is the same as well-foundedness
UbiCC Journal – Volume 4 No. 3
of R^ . Then, I(R) is a monotype. In a concrete
setting, I(R) is the set of monotypes which are not
the origins of infinite paths (by R):
A relation R is progressively finite iff for a mono-
type a, a ≤ (R ◦ a)< ⇒ a = 0 equivalently
ν(a : a ≤ id : (R ◦ a)< ) = 0 equivalently µ(a : a ≤
id : a/•R) = id.
The next theorem involves the function wa (X) :=
Q ∨ P ◦ X, which is closely related to the description
of iterations. The theorem highlights the importance
of progressive finiteness in the simplification of fixed
point-related properties.
(6) Theorem. Let f (X) := Q ∨ P ◦ X be a func-
tion. If P is progressively finite, the function f has a
unique fixed point which means that ν(f ) = µ(f ) =
P ∗ ◦ Q [1]:
As the demonic calculus will serve as an algebraic
apparatus for defining the denotational semantics
of the nondeterministic programs, we will define in
what follows these operators.
3 Demonic refinement order-
ing
We now define the refinement ordering (demonic in-
clusion) we will be using in the sequel. This ordering
induces a complete join semilattice, called a demonic
semilattice. The associated operations are demonic
join (t), demonic meet (u) and demonic composition
of these operations, and illustrate them with simple
examples. For more details on relational demonic
semantics and demonic operators, see [5, 8, 6, 7, 14].
(7) Definition. We say that a relation Q refines a
relation R [23], denoted by Q v R, iff R< ◦ Q ≤
R and R< ≤ Q< .
(8) Proposition. Let Q and R be relations, then
(a) The greatest lower (wrt v) of Q and R is,
Q t R = Q< ◦ R< ◦ (Q ∨ R),
If Q< = R< then we have t and ∨ coincide
i.e Q t R = Q ∨ R.
573
Special Issue on ICIT 2009 Conference - Applied Computing

(b) If Q and R satisfy the condition Q< ∧ R< =
(Q ∧ R)< , their least upper bound is Q u R =
Q ∧ R ∨ Q≺ ◦ R ∨ R≺ ◦ Q, otherwise, the least
upper bound does not exist. If Q< ∧ R< = 0
then we have u and ∧ coincide i.e Q u R =
Q ∧ R.
For the proofs see [9, 14].
(9) Definition. The demonic composition of rela-
tions Q and R [5] is Q 2 R = (R< /•Q) ◦ Q ◦ R.
In what follows we present some properties of 2 .
(10) Theorem.
(a) (P 2 Q) 2 R = P 2 (Q 2 R),
(b) R total ⇒ Q 2 R = Q ◦ R,
(c) Q function ⇒ Q 2 R = Q ◦ R.
See [5, 6, 7, 14, 35].
Monotypes have very simple and convenient prop-
erties. Some of them are presented in the following
proposition.
(11) Proposition. Let a and b be monotypes. We
have
(a) a = a^ = a2 ,
(a) S(R) = I(P ) ◦ [(P ∨ Q)< /•P ∗ ] ◦ P ∗ ◦ Q., with
the restriction
(b) P < ∧ Q< = 0
Our goal is to show that the operational semantics
a is equal to the denotational one which is given as
the greatest fixed point of the semantic function Q ∨
P 2 X in the demonic semilattice. In other words,
we have to prove the next equation:
F
(a) S(R) = {X|X v Q ∨ P 2 X};
by taking P := t 2 B and Q := t∼ , one gets the
demonic semantics we have assigned to while loops
in previous papers [14, 35]. Other similar definitions
of while loops can be found in [19, 25, 29].
Let us introduce the following abbreviations:
(12) Abbreviation. Let P , Q and X be relations
subject to the restriction P < ∧ Q< = 0 (b) and x
a monotype. The Abbreviations wd , wa , w< , a and l
are defined as follows:
wd (X) := Q ∨ P 2 X,
a := (P ∨ Q)< /•P ∗ ,
wa (X) := Q ∨ P ◦ X,
l := I(P ).
w< (x) := Q< ∨ (P 2 x)< = Q ∨ (P 2 x)<

(b) a 2 b = a ∧ b = b 2 a, (Mnemonics: the subscripts a and d stand for angelic

and demonic, respectively; the subscript < refers to
(c) a ∨ a∼ = id and a ∧ a∼ = 0, the fact that w< is obtained from wd by composi-
tion with <; the monotype a stands for abnormal,
(d) a ≤ b ⇔ b∼ ≤ a∼ ,
since it represents states from which abnormal ter-
(e) a∼ 2 b∼ = (a ∨ b)∼ , mination is not possible; finally, l stands for loop,
since it represents states from which no infinite loop
(f ) (a ∧ b)∼ = (a 2 b)∼ = a∼ ∨ b∼ , is possible.)
(g) a 2 b∼ ∨ b = a ∨ b, In what follows we will be concerned about the
fixed point of wa , w< and wd .
(h) a ≤ b ⇔ a 2 1 ≤ b 2 1.
(13) Theorem. Every fixed point Y of wa (Abbre-
In previous papers [14, 13, 31, 35], we found the viation 12) verifies P ∗ ◦ Q ≤ Y ≤ P ∗ ◦ Q ∨ l∼ 2 1,
semantics of the while loop given by the following and the bounds are tight (i.e. the extremal values are
P

  fixed points).
graph: - e - s -
  The next lemma investigates the relationship be-
Q
tween fixed points of w< and those of wd (cf. Abbre-
viation 12).

UbiCC Journal – Volume 4 No. 3 574

Special Issue on ICIT 2009 Conference - Applied Computing
(14) Lemma. Let h(X) := (P ∨ Q)≺ ∨ (P ◦ X)< and
h1 (x) := (P ∨ Q)≺ 2 1 ∨ P ◦ x.
(a) Y = wd (Y ) ⇒ w< (Y < ) = Y < ,
(b) w< (Y < ) = Y < ⇒ h(Y ≺ ) = Y ≺ ,
(c) h(Y ≺ ) = Y ≺ ⇒ h1 (Y ≺ 2 1) = Y ≺ 2 1,
(15) Lemma. Let Y be a fixed point of wd and b be
a fixed point of w< (Abbreviation 12). The relation
b 2 Y is a fixed point of wd .
(16) Lemma. If Y and Y 0 are two fixed points of
wd (Abbreviation 12) such that Y < = Y 0< and Y < ◦P
is progressively finite, then Y = Y 0 .
The next theorem characterizes the domain of the
greatest fixed point, wrt v, of function wd . This
domain is the set of points for which normal ter-
mination is guaranteed (no possibility of abnormal
termination or infinite loop).
(17) Theorem. Let W be the greatest fixed point,
wrt to v, of wd (Abbreviation 12). We have W < =
a 2 l.
The following theorem is a generalization to a non-
deterministic context of the while statement verifi-
cation rule of Mills [24]. It shows that the greatest
fixed point W of wd is uniquely characterized by con-
ditions (a) and (b), that is, by the fact that W is a
fixed point of wd and by the fact that no infinite loop
is possible when the execution is started in a state
that belongs to the domain of W . Note that we also
have W < ≤ a (see Theorem 17), but this condition
is implicitly enforced by condition (a). Half of this
theorem (the ⇐ direction) is also proved by Sekerin-
ski (the main iteration theorem [29]) in a predicative
programming set-up.
(18) Theorem. A relation W is the greatest fixed
point, wrt v, of function wd (Abbreviation 12), iff
the following two conditions hold:
(a) W = wd (W ),
(b) W < ≤ l.
In what follows we give some applications of our
results.
UbiCC Journal – Volume 4 No. 3
4 Application
In [6, 7], Berghammer and Schmidt propose abstract
relation algebra as a practical means for the specifi-
cation of data types and programs. Often, in these
specifications, a relation is characterized as a fixed
point of some function. Can demonic operators be
used in the definition of such a function? Let us now
show with a simple example that the concepts pre-
sented in this paper give useful insights for answering
this question.
In [6, 7], it is shown that the natural numbers can
be characterized by the relations z and S (zero and
successeur ) the laws
(a) Ø 6= z = zL ∧ zz ^ ⊆ I (z is a point),
SS ^ = I ∧ S^S ⊆
I (S is a one to one application.),
Sz = Ø (z has T a predecessor),
L = {x|z ∪ S ^ x =
x} (generation principle).
For the rest of this section, assume that we are
given a relation algebra satisfying these laws. In this
algebra, because of the last axiom, the inequation
(a) z ∪ S ^ X ⊆ X
obviously has a unique solution for X, namely, X =
L. Because the functiong(X) := z ∪ S ^ X is ∪-
continuous, this solution can be expressed as
(a) L = n≥0 g n (Ø) = n≥0 S ^n z,
S S
where g 0 (Ø) = Ø, g n+1 (Ø) = g(g n (Ø)), S ^0 = I
and S ^n+1 = S ^ S ^n . However, it is shown in [6, 7]
that z t S ^ 2 X ⊆ X, obtained by replacing the
join and composition operators in a by their demonic
counterparts, has infinitely many solutions. Indeed,
from Sz = Ø and the Schröder rule, it follows that
(a) z ∩ S ^ L = Ø,
so that, by definition of demonic join (8(a))
and demonic composition (9), z t S ^ 2 X = (z ∪
S ^ 2 X) ∩ z ∩ (S ^ 2 X)L ⊆ z ∩ S ^ L = Ø. Hence,
any relation R is a solution to z t S ^ 2 X ⊆ X.
Looking at previous papers [14, 32, 33, 34, 31], one
575
Special Issue on ICIT 2009 Conference - Applied Computing
immediately sees why it is impossible to reach L by
joining anything to z (which is a point and hence is
an immediate predecessor of Ø), since this can only
lead to z or to Ø.
Let us now go ‘fully demonic’ and ask what is a
solution to ztS ^ 2 X v X. By the discussion above,
this is equivalent to Ø v X, which has a unique
solution, X = Ø. This raises the question whether
it is possible to find some fully demonic inequation
similar to (a), whose solution is X = L. Because L is
in the middle of the demonic semilattice, there are in
fact two possibilities: either approach L from above
or from below.
For the approach from above, consider the inequa-
tion
X v z u S ^ 2 X.
Using Theorem 10(c), we have z u S ^ 2 X =
z u S ^ X, since S ^ is deterministic (axiom a(b)).
From a, z ⊆ S ^ L; this implies z ⊆ S ^ XL and
S ^ X ⊆ z, so that, by definition of u,
z u S ^ X = z ∩ S ^ X ∪ z ∩ S ^ XL ∪ z ∩ S ^ X =
z ∪ S ^ X.
This means that 4 reduces to
(a) X v z ∪ S ^ X.
By definition of refinement (7), this implies that
z ∪ S ^ XL ⊆ XL; this is a variant of (a), thus
having XL = L as only solution. This means that
any solution to 4 must be a total relation. But L
is total and in fact is the largest (by v) total rela-
tion. It is also a solution toF4 (since by axiom a(d),
z ∪ S ^ L = L) so that L = {X|X v z u S ^ 2 X};
that is, L is the greatest fixed point in (BL , v) of
f (X) := z u S ^ 2 X. Now consider n≥0 S
^n 2
z,
^n
where S is a n-fold demonic composition defined
by S ^0 = I and S ^n+1 = S ^ 2 S ^n . By axiom
a(b), S ^ is deterministic, so that, by 10(c) and asso-
ciativity of demonic composition, conS n 2 z = S ^n z.
Hence,
It is easy to show that for any n ≥ 0, S ^n z is
a point (it is the n-th successor of zero) and that
m 6= n ⇒ S ^m z 6= S ^n z. Hence, in (BL , v),
{S ^n z|n ≥ 0} (i.e. {S ^n 2 z|n ≥ 0}) is the set of
immediate predecessors of Ø; looking at [31] shows
UbiCC Journal – Volume 4 No. 3
how the universal relationL arises as the greatest
lower bound n≥0 S ^n 2 z of this set of points. Note
that, whereas there is a unique solution to a, there
are infinitelyF
many solutionsSto 4 (equivalently, to a),
for example n≥k S n (= n≥k S n ), for any k.
For the upward approach, consider
z ^ t X 2 S v X.
Here also there are infinitely many solutions to this
inequation; in particular, any vector v, including
Ø and L, is a solution to 4. Because (BL , v) is
only a join semilattice, it is not at all obvious that
the least fixed point of h(X) := z ^ t X 2 S ex-
ists. It does, however, since the following deriva-
tion shows that n≥0 z ^ 2 S n (=
F F
h n ^
(z ),
n≥0
0 ^ ^
where h (z ) = z ) is a fixed point of h and hence
is obviously the least solution of 4: Because z ^
and S are mappings, property 10(c) implies that
z ^ 2 S n = z ^ S n , for any n ≥ 0. But z ^ S n is
also a mapping (it is the inverse of the point S ^n z)
and hence is total,Ffrom which, by Proposition 8(a)
^2 n ^ n
F
and equation a, z
S n≥0 ^n S = n≥0 z S =
^ n ^
S
n≥0 z S = ( n≥0 S z)˘ = L = L. This
means that L is the least upper bound of the set
of mappings {z ^ 2 S n |n ≥ 0}. Again, a look at
[31] gives some intuition to understand this result,
after recalling that mappings are minimal elements
in (BL , v) (though not all mappings have the form
z ^ 2 S n ).
Thus, building L from below using the set of map-
pings {z ^ 2 S n |n ≥ 0} is symmetric to building it
from above using the set of points {S ^n 2 z|n ≥ 0}.
5 Conclusion
We presented a theorem that can be also used to find
the fixed points of functions of the form f (X) :=
Q ∨ P 2 X (no restriction on the domains of P and
Q). This theorem can be applied also to the program
verification and construction (as in the precedent ex-
ample). Half of this theorem (the ⇐ direction) is
also proved by Sekerinski (the main iteration theo-
rem [29]) in a predicative programming set-up. Our
theorem is more general because there is no restric-
tion on the domains of the relations P and Q.
576
Special Issue on ICIT 2009 Conference - Applied Computing
The approach to demonic input-output relation
presented here is not the only possible one. In
[19, 20, 21], the infinite looping has been treated by
adding to the state space a fictitious state ⊥ to de-
note nontermination. In [8, 18, 22, 26], the demonic
input-output relation is given as a pair (relation,set).
The relation describes the input-output behavior of
the program, whereas the set component represents
the domain of guaranteed termination.
We note that the preponderant formalism em-
ployed until now for the description of demonic
input-output relation is the wp-calculus. For more
details see [3, 4, 17].
References
[1] Backhouse, R. C., and Doombos, H.: Math-
ematical Induction Made Calculational. Com-
puting science note 94/16, Department of Math-
ematics and Computer Science, Eindhoven Uni-
versity of Technology, The Netherlands, 1994.
[2] Backhouse, R. C., Hoogendijk, P., Voermans,
E. and van der Woude, J.:. A Relational The-
ory of Datatypes. Research report, Department
of Mathematics and Computer Science, Eind-
hoven University of Technology, The Nether-
lands, 1992.
[3] R. J. R. Back. : On the correctness of refinement
in program development. Thesis, Department of
Computer Science, University of Helsinki, 1978.
[4] R. J. R. Back and J. von Wright.: Combining
angels, demons and miracles in program spec-
ifications. Theoretical Computer Science,100,
1992, 365–383.
[5] Backhouse, R. C. and van der Woude, J.: De-
monic Operators and Monotype Factors. Math-
ematical Structures in Comput. Sci., 3(4), 417–
433, Dec. (1993). Also: Computing Science Note
92/11, Department of Mathematics and Com-
puter Science, Eindhoven University of Technol-
ogy, The Netherlands, 1992.
UbiCC Journal – Volume 4 No. 3
[6] Berghammer, R.: Relational Specification of
Data Types and Programs. Technical report
9109, Fakultät für Informatik, Universität der
Bundeswehr München, Germany, Sept. 1991.
[7] Berghammer, R. and Schmidt, G.: Relational
Specifications. In C. Rauszer, editor, Algebraic
Logic, 28 of Banach Center Publications. Polish
Academy of Sciences, 1993.
[8] Berghammer, R. and Zierer, H.: Relational Al-
gebraic Semantics of Deterministic and Nonde-
terministic Programs. Theoretical Comput. Sci.,
43, 123–147 (1986).
[9] Boudriga, N., Elloumi, F. and Mili, A.: On
the Lattice of Specifications: Applications to a
Specification Methodology. Formal Aspects of
Computing, 4, 544–571 (1992).
[10] Chin, L. H. and Tarski, A.: Distributive and
Modular Laws in the Arithmetic of Relation Al-
gebras. University of California Publications, 1,
341–384 (1951).
[11] Conway, J. H.: Regular Algebra and Finite Ma-
chines. Chapman and Hall, London, 1971.
[12] Davey, B. A. and Priestley, H. A.: Introduction
to Lattices and Order. Cambridge Mathematical
Textbooks. Cambridge University Press, Cam-
bridge, 1990.
[13] J. Desharnais, B. Möller, and F. Tchier. Kleene
under a demonic star. 8th International Con-
ference on Algebraic Methodology And Software
Technology (AMAST 2000), May 2000, Iowa
City, Iowa, USA, Lecture Notes in Computer
Science, Vol. 1816, pages 355–370, Springer-
Verlag, 2000.
[14] Desharnais, J., Belkhiter, N., Ben Mo-
hamed Sghaier, S., Tchier, F., Jaoua, A., Mili,
A. and Zaguia, N.: Embedding a Demonic Semi-
lattice in a Relation Algebra. Theoretical Com-
puter Science, 149(2):333–360, 1995.
577
Special Issue on ICIT 2009 Conference - Applied Computing
[15] Desharnais, J., Jaoua, A., Mili, F., Boudriga,
N. and Mili, A.: A Relational Division Oper-
ator: The Conjugate Kernel. Theoretical Com-
put. Sci., 114, 247–272 (1993).
[16] Dilworth, R. P.: Non-commutative Residuated
Lattices. Trans. Amer. Math. Sci., 46, 426–444
(1939).
[17] E. W. Dijkstra. : A Discipline of Programming.
Prentice-Hall, Englewood Cliffs, N.J., 1976.
[18] H. Doornbos. : A relational model of programs
without the restriction to Egli-Milner monotone
constructs. IFIP Transactions, A-56:363–382.
North-Holland, 1994.
[19] C. A. R. Hoare and J. He. : The weakest
prespecification. Fundamenta Informaticae IX,
1986, Part I: 51–84, 1986.
[20] C. A. R. Hoare and J. He. : The weakest
prespecification. Fundamenta Informaticae IX,
1986, Part II: 217–252, 1986.
[21] C. A. R. Hoare and al. : Laws of programming.
Communications of the ACM, 30:672–686, 1986.
[22] R. D. Maddux. : Relation-algebraic semantics.
Theoretical Computer Science, 160:1–85, 1996.
[23] Mili, A., Desharnais, J. and Mili, F.: Relational
Heuristics for the Design of Deterministic Pro-
grams. Acta Inf., 24(3), 239–276 (1987).
[24] Mills, H. D., Basili, V. R., Gannon, J. D. and
Hamlet,R. G.: Principles of Computer Pro-
gramming. A Mathematical Approach. Allyn
and Bacon, Inc., 1987.
[25] Nguyen, T. T.: A Relational Model of Demonic
Nondeterministic Programs. Int. J. Founda-
tions Comput. Sci., 2(2), 101–131 (1991).
[26] D. L. Parnas. A Generalized Control Structure
and its Formal Definition. Communications of
the ACM, 26:572–581, 1983
[27] Schmidt, G.: Programs as Partial Graphs I:
Flow Equivalence and Correctness. Theoretical
Comput. Sci., 15, 1–25 (1981).
UbiCC Journal – Volume 4 No. 3
[28] Schmidt, G. and Ströhlein, T.: Relations and
Graphs. EATCS Monographs in Computer Sci-
ence. Springer-Verlag, Berlin, 1993.
[29] Sekerinski, E.: A Calculus for Predicative Pro-
gramming. In R. S. Bird, C. C. Morgan, and
J. C. P. Woodcock, editors, Second Interna-
tional Conference on the Mathematics of Pro-
gram Construction, volume 669 of Lecture Notes
in Comput. Sci. Springer-Verlag, 1993.
[30] Tarski, A.: On the calculus of relations. J.
Symb. Log. 6, 3, 1941, 73–89.
[31] F. Tchier.: Sémantiques relationnelles
démoniaques et vérification de boucles non
déterministes. Theses of doctorat, Département
de Mathématiques et de statistique, Université
Laval, Canada, 1996.
[32] F. Tchier.: Demonic semantics by mono-
types. International Arab conference on In-
formation Technology (Acit2002),University of
Qatar, Qatar, 16-19 December 2002.
[33] F. Tchier.: Demonic relational semantics of
compound diagrams. In: Jules Desharnais,
Marc Frappier and Wendy MacCaull, editors.
Relational Methods in computer Science: The
Québec seminar, pages 117-140, Methods Pub-
lishers 2002.
[34] F. Tchier.: While loop d demonic relational
semantics monotype/residual style. 2003 In-
ternational Conference on Software Engineer-
ing Research and Practice (SERP03), Las Ve-
gas, Nevada, USA, 23-26, June 2003.
[35] F. Tchier.: Demonic Semantics: using mono-
types and residuals. IJMMS 2004:3 (2004) 135-
160. (International Journal of Mathematics and
Mathematical Sciences)
[36] M. Walicki and S. Medal.: Algebraic approches
to nondeterminism: An overview. ACM compu-
tong Surveys,29(1), 1997, 30-81.
[37] L.Xu, M. Takeichi and H. Iwasaki.: Rela-
tional semantics for locally nondeterministic
578
Special Issue on ICIT 2009 Conference - Applied Computing

programs. New Generation Computing 15, 1997,

339-362.

UbiCC Journal – Volume 4 No. 3 579

Special Issue on ICIT 2009 Conference - Applied Computing

CASE STUDIES IN THIN CLIENT ACCEPTANCE

Paul Doyle, Mark Deegan, David Markey, Rose Tinabo, Bossi Masamila, David Tracey
School of Computing, Dublin Institute of Technology, Ireland
WiSAR Lab, Letterkenny Institute of Technology
{paul.doyle, mark.deegan, david.markey}@dit.ie,{rose.tinabo, bossi.masamila}@student.dit.ie
david.tracey@lyit.ie

ABSTRACT
Thin Client technology boasts an impressive range of financial, technical and
administrative benefits. Combined with virtualisation technology, higher
bandwidth availability and cheaper high performance processors, many believe
that Thin Clients have come of age. But despite a growing body of literature
documenting successful Thin Client deployments there remains an undercurrent
of concern regarding user acceptance of this technology and a belief that greater
efforts are required to understand how to integrate Thin Clients into existing,
predominantly PC-based, deployments. It would be more accurate to state that
the challenge facing the acceptance of Thin Clients is a combination of
architectural design and integration strategy rather than a purely technical issue.
Careful selection of services to be offered over Thin Clients is essential to their
acceptance. Through an evolution of three case studies the user acceptance issues
were reviewed and resolved resulting in a 92% acceptance rate of the final Thin
Client deployment. No significant bias was evident in our comparison of user
attitudes towards desktop services delivered over PCs and Thin Clients.

Keywords: Thin Clients, Acceptance, Virtualisation, RDP, Terminal Services.

1 INTRODUCTION technology. Over a four year period, three Thin

It is generally accepted that in 1993 Tim Negris
coined the phrase “Thin Client” in response to Larry
Ellison’s request to differentiate the server centric
model of Oracle from the desktop centric model
prevalent at the time. Since then the technology has
evolved from a concept to a reality with the
introduction of a variety of hardware devices,
network protocols and server centric virtualised
environments. The Thin Client model offers users
the ability to access centralised resources using full
graphical desktops from remotely located, low cost,
stateless devices. While there is sufficient literature
in support of Thin Clients and their deployment, the
strategies employed are not often well documented.
To demonstrate the critical importance of how Thin
Clients perform in relation to user acceptance we
present a series of case studies highlighting key
points to be addressed in order to ensure a successful
deployment.
1.1 Research Aim
The aim of this research has been to identify a
successful strategy for Thin Client acceptance within
an educational institute. There is sufficient literature
which discusses the benefits of Thin Client adoption,
and while this was referenced it was not central to
the aims of this research as the barrier to obtaining
these benefits was seen to be acceptance of the
Client case studies were run within the Dublin
Institute of Technology with the explicit aim of
determining the success factors in obtaining user
satisfaction. The following data criteria were used to
evaluate each case study in addition to referencing
the Universal Theory of User Acceptance Testing
(UTUAT) [1].
1) Login events on the Thin Clients.
2) Reservation of the Thin Client facility.
3) The cost of maintaining the service.
1.2 Paper Structure
In section 2 we review the historical background
and trends of Thin Client technology to provide an
understanding of what the technology entails.
Section 3 discusses the case for Thin Clients within
existing literature including a review of deployments
within industry and other educational institutes.
Section 4 provides details of the three case studies
discussing their design, evaluating the results, and
providing critical analysis. Section 5 takes a critical
look at all of the data and sections 6 and 7 provide
conclusions and identify future work. This paper is
aimed at professionals within educational institutes
seeking ways to realize the benefits of Thin Client
computing while maintaining the support and
acceptance of users. It provides a balance between

UbiCC Journal – Volume 4 No. 3 585

Special Issue on ICIT 2009 Conference - Applied Computing
the hype of Thin Clients and the reality of their
deployment.
2 THIN CLIENT EVOLUTION
The history of Thin Clients is marked by a
number of overly optimistic predictions that it was
about to become the dominant model of desktop
computing. In spite of this there have been a number
of marked developments in this history along with
those of desktop computing in general which are
worth reviewing to set the context for examining the
user acceptance of this technology. Thin Clients have
established a role in desktop computing although not
quite the dominant one initially predicted. These
developments have usually been driven by increases
in processing power (and reductions in the processor
costs) in line with Moore's law, but the
improvements in bandwidth and storage capacity are
having an increasing effect on desktop computing
and on Thin Client computing [2] driving the move
towards more powerful lower cost desktops but also
the possibilities of server virtualisation and Thin
Client computing with the ability to run Thin Clients
over WANs.
The first wave of computing was one where
centralised mainframe computers provided the
computing power as a shared resource which users
accessed using dumb terminals which provided basic
text based input and output and then limited graphics
as they became graphics terminals. These
mainframes were expensive to purchase and were
administered by specialists in managed environments
and mostly used for specific tasks such as
performing scientific calculations and running highly
specialised bespoke payroll systems.
The next wave was that of personal computing,
whereby users administered their own systems which
provided a platform for their personal applications,
such as games, word-processor, mail and personal
data. Since then the personal computer has
undergone a number of significant changes, but the
one of most interest was the nature of the interface
provided to the user which has grown into a rich
Graphical User Interface where the Personal
Computer became a gateway to the Internet with the
Web browser evolving into a platform for delivery of
rich media content, such as audio and video.
This move from a mainframe centralised
computing model to a PC distributed one resulted in
a number of cost issues related to administration.
This issue was of particular concern for corporate
organizations, in relation to licensing, data security,
maintenance and system upgrades. For these cost
reasons and the potential for greater mobility for
users, the use of Thin Clients is often put forward as
a way to reduce costs using the centralised model of
the Thin Client architecture. This also offers lower
purchase costs and reduces the consumption of
energy [3].
UbiCC Journal – Volume 4 No. 3
The challenge faced by Thin Client technology is
to deliver on these lower costs and mobility, while
continuing to provide a similarly rich GUI user
experience to that provided by the desktop machine
(a challenge helped by improved bandwidth, but
latency is still often a limiting factor [4]) and the
flexibility with regard to applications they have on
their desktop. Typically, current Thin Client systems
have an application on a server (generally Windows
or Linux) which encodes the data to be rendered into
a remote display protocol. This encoded data is sent
over a network to a Thin Client application running
on a PC or a dedicated Thin Client device to be
decoded and displayed. The Thin Client will send
user input such as keystrokes to the application on
the server. The key point is that the Thin Client does
not run the code for the user's application, but only
the code required to support the remote display
protocol.
While the term Thin Client was not used for
dumb terminals attached to mainframes in the 1970's,
the mainframe model shared many of the attributes
of Thin Client computing. It was centralised, the
mainframe ran the software application and held the
data (or was attached to the data storage) and the
terminal could be shared by users as it did not retain
personal data or applications, but displayed content
on the screen as sent to it by the mainframe. From a
desktop point of view, the 1980's were dominated by
the introduction and adoption of the Personal
Computer.
Other users requiring higher performance and
graphics used Unix Workstations from companies
like Apollo and Sun Microsystems. The X Window
System [5] was used on many Workstations and X
terminals were developed as a display and input
terminal and provided a lower cost alternative to a
Unix Workstation, with the X terminal connecting to
a central machine running an X display manager. As
such, they shared some of the characteristics of a
Thin Client system, although the X terminal ran an X
Server making it more complicated than Thin Client
devices.
The 1990's saw the introduction of several remote
display protocols, such as Citrix's ICA [6]
Microsoft's RDP [7] and AT&T's VNC [8] for Unix
that took advantage of the increasing bandwidth
available on a LAN to provide a remote desktop to
users.
Terminal Services was introduced as part of
Windows NT4.0 in 1996 and it offered support for
the Remote Desktop Protocol (RDP) allowing access
to Windows applications running on the Server,
giving users access to a desktop on the Server using
an RDP client on their PC. RDP is now offered on a
range of Windows platforms [9]. Wyse and vendors
such as Ncomputing launched terminals, which
didn't run the Windows operating system, but
accessed Windows applications on a Windows
Server using RDP, which is probably still the
586
Special Issue on ICIT 2009 Conference - Applied Computing
dominant role of dedicated hardware Thin Clients.
Similarly VNC is available on many Linux and Unix
distributions and is commonly used to provide
remote access to a user's desktop. These remote
display protocols face increasing demands for more
desktop functionality and richer media content, with
ongoing work required in how, where and when
display updates are encoded, compressed or cached
[10]. Newer remote display protocols such as THINC
have been designed with the aim of improving these
capabilities [11].
In 1999, Sun Microsystems took the Thin Client
model further with the SunRay, which was a simple
network appliance, using its own remote display
protocol called ALP. Unlike some of the other Thin
Clients which ran their own operating system,
SunRay emphasized its completely stateless nature
[12]. This stateless nature meant that no session
information or data was held or even cached (not
even fonts) on the appliance itself and enabled its
session mobility feature, whereby a smart card was
used to identify a user with a session so that with the
smartcard the user could login from any SunRay
connected to the session's server and receive the
desktop as it was previously.
Many of these existing players have since
focused on improving their remote desktop protocols
and support for multimedia or creating new hardware
platforms. There have also been some newer arrivals
like Pano Logic and Teradici who have developed
specific client hardware to create “zero” clients, with
supporting server virtualisation to render the remote
display protocols. Also, there are a number of
managed virtual desktops hosted in a data centre now
being offered.
One of the drivers behind Thin Client
Technology, particularly when combined with a
dedicated hardware device, is to reduce the cost of
the client by reducing the processing requirement to
that of simply rendering content, but a second driver
(and arguably more important one) is to gain a level
of universality by simplifying the variations in the
client side environment. This has been met in a
number of new ways using Virtual Machine players
and USB memory in Microsoft's research project
“Desktop on a Keychain” (DOK) [13] and also the
Moka5 product [14], allowing the mobility (and
security) benefits attributed to Thin Clients. This can
be enhanced with the use of network storage to cache
session information [15].
It can be seen that Thin Clients have evolved
along with other desktop computing approaches,
often driven by the same factors of increasing
processing power, storage capacity and bandwidth.
However, newer trends that are emerging with regard
to virtualisation, internet and browser technologies,
together with local storage, present new challenges
and opportunities for Thin Client technology to win
user acceptance. As Weiser said in 1999 in this new
era, “hundreds or thousands of computers do our
UbiCC Journal – Volume 4 No. 3
bidding. The relationship is the inverse of the
mainframe era: the people get the air conditioning
now, and the nice floors, and the computers live out
in cyberspace and sit there waiting eagerly to do
something for us”. [16]
3 THE CASE FOR THIN CLIENTS
There are many stated benefits for Thin Clients
all of which are well documented [17][18]. While
there is no single definitive list, potential system
designers may have different aims when considering
Thin Clients, these benefits should be clearly
understood prior to embarking on any deployment
and are discussed below.
3.1 Reduced cost of software maintenance
The administrative cost benefit of the Thin
Client model, according to Jern [19] is based on the
simple observation that there are fewer desktop
images to manage. With the combination of
virtualisation environments and Windows Terminal
Service (WTS) systems it would not be uncommon
for twenty five or more desktop environments to be
supported from a single installation and
configuration. This reduces the number of upgrades
and customizations required for desktop images in
computer laboratories where the aim is to provide a
consistent service from all systems. Kissler and Hoyt
[20] remind us that the “creative use of Thin Client
technology can decrease both management
complexity and IT staff time.” In particular they
chose Thin Client technology to reduce the
complexity of managing a large number of kiosks
and quick-access stations in their new thirty three
million dollar library. They have also deployed Thin
Client devices in a range of other roles throughout
Valparaiso University in Indiana. Golick [21] on the
other hand suggests that the potential benefits of a
Thin Client approach include the lower mean time to
repair (MTTR) and lower distribution costs. It is
interesting to note that he does suggest that the
potential cost savings for hardware are a myth, but
that administration savings still make a compelling
case for using Thin Client technology.
3.2 Enhanced Security
Speer and Angelucci [22] suggest that security
concerns should be a major factor in the decision to
adopt Thin Client systems and this becomes more
apparent when referencing the Gartner Thin Client
classification model. The Thin Client approach
ensures that data is stored and controlled at the data-
centre hosting the Thin Client devices. It is easy to
argue that the user can retain the mobility of laptops
but with enhanced security and the data is not
mobile, just the access point. The argument is even
easier to make when we consider recent high-profile
cases of the theft of unencrypted laptops containing
sensitive medical or financial records. The freedom
587
Special Issue on ICIT 2009 Conference - Applied Computing
conferred on users of corporate desktop and laptop
PCs undermines the corporation’s obligations in
relation to data privacy and security. Steps taken to
protect sensitive data on user devices are often too
little and too late. Strassmann [23] states that the
most frequent use of a personal computer is for
accessing web applications and states that the Thin
Client model demonstrates significantly lower
security risks for the corporation. Five security
justifications for adopting the Thin Client model
were proposed.
1) Zombie Prevention
2) Theft Dodging
3) File Management
4) Software Control
5) Personal Use Limitations
Strassmann concedes that Thin Clients are not
necessarily best for every enterprise and every class
of user, but for enterprises with a large number of
stationary “non-power” users, “Thin Clients may
present the best option in terms of security, cost
effectiveness and ease of management.”
3.3 User Mobility
User mobility can refer to the ability of a user to
use any device, typically within the corporation’s
intranet, as a desktop where the user will see a
consistent view of the system, for example, SunRay
hot-desking. While user profiles in Microsoft
Windows support this, it is often only partially
implemented. Session mobility can be viewed as the
facility for users to temporarily suspend or
disconnect their desktop session and to have it re-
appear, at their request, on a different device at a
later time. This facility removes the need for users to
log-out or to boot-up a desktop system each time
they wish to log-in. Both of these potential features
of Thin Client technologies help to break the sense of
personal ownership that users often feel for their
desktop or laptop computers. It is this sense of
personal ownership which makes the maintenance
and replacement of corporate PCs a difficult task,
and this feeling of ownership and control is often a
reason why users resist the adoption of a centrally
controlled Thin Client to replace their desktop,
whereas this is exactly why IT management may
want to adopt it.
3.4 Environmental Costs
In the article “An Inefficient Truth” Plan [24]
reveals a series of “truths” supported by a number of
case studies directed at the growing costs of
Information and Communication Technologies. One
such case study is of Reed Managed Services where
4,500 PCs were replaced with Thin Clients, and a
centralised blade server providing server based
virtualised desktops. Savings are reported as follows:
UbiCC Journal – Volume 4 No. 3
1) 5.4 million kWh reduction,
2) 2,800 tonnes of CO2 saved annually
3) Servers reduced by a factor of 20
4) IT budget cut by a fifth
Indeed there are many deployments focused on
obtaining energy savings through the use of Thin
Clients. In a case study where SunRay systems were
introduced into Sparkasse a public German Bank,
Bruno-Britz [25] reports that the savings in
electricity costs alone were enormous. The
University of Oxford has deployed SunRay Thin
Client devices in their libraries citing the cooler and
quieter operation as factors in their decision. These
devices, having no local hard disk and no fan operate
at a lower temperature and more quietly than
traditional PCs. This characteristic has
environmental implications from noise, cooling and
power consumption perspectives.
3.5 Summary of Benefits
In summary, we can extract the benefits
observed within literature and case studies as
follows:
1) Increased security as data maintained centrally
2) Reduced cost of hardware deployment and
management and faster MTTR
3) Reduced administration support costs
4) Environmental costs savings
5) Reduced cost of software maintenance
6) Reduced cost of software distribution
7) Zero cost of local software support
8) The ability to leverage existing desktop hardware
and software
9) Interface portability and session mobility
10) Enhanced Capacity planning
11) Centralised Usage Tracking and Capacity
Planning
3.6 Thin Clients vs. Fat Clients
Thin Client technology has evolved in
sophistication and capability since the middle of the
1990s, however the “thickness” (the amount of
software and administration required on the access
device) of the client is a source of distinction for
many vendors [26][27]. Regardless of “thickness”,
Thin Clients require less configuration and support
when compared to Fat Clients (your typical PC). In
the early 1990s Gartner provided a client-server
reference design shown in Figure 1. This design
provides clarity for the terms “thin” and “fat” clients
by viewing applications in terms of the degree of
data access, application and presentation logic
present on the server and client sides of the network.
The demand for network based services such as
email, social networking and the World Wide Web
has driven bandwidth and connectivity requirements
to higher and higher levels of reliability and
performance [28]. As we progress to an “always on”
588
Special Issue on ICIT 2009 Conference - Applied Computing
network infrastructure the arguments focused against
Thin Clients based on requiring an offline mode of
usage are less relevant. The move from Fat Client to
Thin Client is however often resisted as individuals
find themselves uncomfortable with the lack of
choice provided when the transition is made, as
observed by Wong et al.[29].
Figure 1: Gartner Group Client/Server Reference Design
4 CASE STUDIES
No matter how well documented the benefits of
Thin Clients may be, there is still an issue of
acceptance to be addressed. While it may be
tempting to assume that the implementation of
technology is a technical issue and that simply by
building solutions a problem is effectively solved,
evidence would point to the contrary. As there can
often be a disparity between what is built and what is
required or needed. Too often requirements
gathering, specification definition and user
consultation are forgotten in the rush to provide new
services which are believed to be essential. In
essence the notion of “if we build it they will come”
is adopted, inevitably causing confusion and
frustration for both service provider and the user. For
example, during Sun Microsystems’ internal
deployment of its own SunRay Thin Client solution
many groups and functions sought exemptions from
the deployment as they believed that their
requirements were sufficiently different to the
“generic user” to warrant exclusion from the project.
The same arguments still exist today and it is often
those with a more technical understanding of the
technology who are the agents of that technology’s
demise. By providing interesting and often creative
edge cases which identify the limitations of a
technology, they can, by implication, tarnish it as an
UbiCC Journal – Volume 4 No. 3
incomplete and flawed technology. In the case of
Thin Clients, it should be accepted that there are
tradeoffs to be made. One of the appealing aspects of
the Fat client is its ability to be highly flexible which
facilitates extensive customization. However not
every user will require that flexibility and
customization. Thin Clients are not going to be a
silver bullet addressing all users needs all of the
time.
All three case studies were evaluated under the
following headings in order to allow a direct
comparison between each. These criteria were
selected to ensure that there was a balance between
the user acceptance of the technology and the
technical success of each deployment.
1) Login events on the Thin Clients
2) Reservation of the Thin Client facility
3) The cost of maintaining the service
Figure 2: Case Study 1
4.1 DIT Case Study 1
In 2005 the DIT introduced the SunRay Thin
Client technology into the School of Computing. In a
similar approach to many other technology
deployments the strengths of the technology were
reviewed and seen as the major selling points of the
deployment. In the case of SunRay there was a cheap
appliance available which would provide the service
of graphical based Unix desktops. Centralised
administration ensured that the support costs would
be low and the replacement requirements for systems
for the next five years would be negligible. In
essence the technological and administrative
advantages were the focus of this deployment. Few
of the services offered within the existing PC
infrastructure were included in the deployment. This
deployment sought to offer new services to students
and introduced Thin Clients for the first time to both
students and staff.
589
Special Issue on ICIT 2009 Conference - Applied Computing
4.1.1 Design
A single laboratory was identified for deploying
the SunRay systems and all PC in that lab were
replaced with SunRay 150 devices. A private
network interconnect was built which ensured that all
data sent from the clients traversed a private network
to the SunRay server. The initial design of this case
study is shown in Figure 2 and it allowed students
within this new Thin Client lab access to the latest
version of Solaris using a full screen graphical
environment as opposed to an SSH command-line
Unix shell which was the traditional method still
used from existing computing laboratories. A new
authentication system was introduced based on
LDAP which required students to have a new
username and password combination which was
different to the credentials already in use within the
Active Directory domain used for the existing PC
network. The reason for this alternative
authentication process was due to the difficulty of
authenticating on a Unix system using Active
Directory. Once the server was running, the Thin
Client laboratory was ready to provide graphical
based Unix login sessions at a considerable reduced
price when compared to an investment of Unix
workstations for each desk. In total 25 Thin Client
devices were installed which were all connected to a
single Solaris server. In summary the key
components within the design were as follows:
1) The service was on a private network
2) New authentication process was introduced
3) New storage mechanism was introduced
4) Devices were all in the same location
5) Service provided was a CDE desktop on Solaris
6) Graphical desktops running on Linux servers also
accessible
4.1.2 Results
The login events are a measure of the general
activity of the devices themselves and were
considered to be a reasonable benchmark for
comparison with existing laboratories within the
institute. One interesting point is that the comparison
of facilities is not necessarily relevant when the
facilities provide different services. Due to the fact
that Unix instead of Windows was provided meant
that, with the exception of those taking courses
involving Unix, the majority of students were
unfamiliar with the technology and did not seek to
use the systems.
Login events on the Thin Clients:
The login events were extracted from the Solaris
server by parsing the output of the last command
which displays the login and logout information for
users which it extracts from the /var/adm/wtrmpx
file. The number of login events per day was
calculated and plotted in the graph shown in Fig. 3.
Immediately obvious was the low use of the system.
UbiCC Journal – Volume 4 No. 3
Given that the nature of the service did not
significantly change over the course of the three
years that the system was in place with the exception
of semester activity in line with student presence in
the institute, it is clear that there was low utilization
of the service. The graph shows raw data plotted,
where login events were less than 10 per day.
14
12
10
Login Events per day
8
6
4
2
0
Feb 05 Feb 06 Feb 07 Feb 08
Figure 3: User Login Events
Reservation of the Thin Client Facility:
Each laboratory may be reserved by staff for the
delivery of tutorial sessions and exercises. The
hourly reservations for this laboratory were reduced
as a result of the introduction of Thin Clients with
only 1 to 2 hours being reserved per day. One of the
primary reasons for the reduction in the use of this
facility was the fact that it had now become special
purpose and the bookings for the room were limited
to the courses which could be taught within it.
The Cost of Maintaining the Service:
A detailed analysis of cost savings associated with
the introduction of Thin Clients within our institute
and specifically the costs associated with this case
study was performed by Reynolds and Gleeson, [30].
In their study they presented evidence of savings in
relation to the cost of support, the cost of deployment
and a basic analysis of the power consumption costs.
They review both the system and the software
distribution steps associated with Thin Clients and
PC systems and present a point of quantifiable
comparison between the two. Key findings of this
analysis were as follows:
1) Time spent performing system upgrades and
hardware maintenance was reduced to virtually
zero as no hardware or software upgrades were
required.
2) A single software image was maintained at the
central server location and changes were made
available instantly to all users.
3) No upgrade costs were incurred on the Thin
Clients or server hardware. All systems have
590
Special Issue on ICIT 2009 Conference - Applied Computing

remained in place throughout both case studies. This is defined as the degree to which there is a
The devices in this lab are now 8 years old and perception of how others will view or judge them
are fulfilling the same role today as they did based on their use of the system. Clearly by
when first installed. isolating the devices and having it associated
4) The Thin Client lab is a low power consumption with specialized courses, there was no social
environment due to the inherent energy efficiency imperative to use the labs. Unix as a desktop was
of the Thin Client hardware over existing PCs. relatively uncommon in the School at the time of
This can provide up to 95% energy savings when the case study and there would have been a
compared to traditional PCs [24]. moderate to strong elitist view of those who were
technical enough to use the systems.
4.1.3 Analysis
d) Facilitating Conditions
There has been extensive research in the area of
This is defined as the degree to which an
user acceptance of technology, but perhaps the most
individual believes in the support for a system. At
relevant work in this area is the Unified Theory of
first glance this does not appear to be a
Acceptance and Use of Technology (UTAUT) [1]
significant factor considering that the services
which identifies four primary constructs or factors;
were created by the support team and there was
considerable vested interest in seeing it succeed.
a) Performance Expectancy
However additional questions asked by the
b) Effort Expectancy
UTAUT include the issue of compatibility with
c) Social Influence
systems primarily used by the individual.
d) Facilitating Conditions
Each of the UTAUT factors can be considered
While there are additional factors such as
significant for Case Study 1. Many of the issues
Gender, Age and Experience, within the student
raised hang on the fundamental issue that the new
populations these are for the most part reasonably
services offered on the Thin Client were different to
consistent and will be ignored. It should be stressed
existing services and for all practical purposes seen
that although the UTAUT was developed for an
as incompatible with the majority of systems
industry based environment it is easily adapted for
available to students elsewhere. The fact that the
our purposes. It was felt that this model serves as a
technology itself may have worked flawlessly, and
relevant reference point when discussing the
may have delivered reduced costs was irrelevant as
performance of the case studies.
the service remained under utilized. Given that the
Clearly Case Study 1 failed to gain acceptance
reason for this lack of acceptance was potentially
despite belief that it would in fact be highly
inherent in the implementation of services and not
successful at its inception. We review the case study
due to failings in the technology itself it was clear
under the four UTAUT headings to identify the
that a second case study was required which would
source of the user rejection of the Thin Clients.
address the issue of service.
a) Performance Expectancy
This factor is concerned with the degree to which
the technology will assist in enhancing a users
own performance. Clearly however the services
provided an advantage to those students who
wished to use Unix systems. Since the majority
of courses are based on the Windows operating
system it would be reasonable to assume that
there was no perceived advantage in using a
system which was not 100% compatible with the
productivity applications used as part of the
majority of courses.
b) Effort Expectancy
This factor is concerned with the degree of ease
associated with the use of the system. One of the
clear outcomes of Case Study 1 was that students
rejected the Unix systems as it was seen to be a
highly complex system, requiring additional
authentication beyond what was currently used in
traditional laboratories. Figure 4: Case Study 2
c) Social Influence

UbiCC Journal – Volume 4 No. 3 591

Special Issue on ICIT 2009 Conference - Applied Computing
4.2 Case Study 2
The second case study is a modification of the
basic implementation of the first case study with
changes focused on increasing student acceptance of
the Thin Client facility. Removing the Unix centric
nature of the existing service was central to the
system redesign. It was decided that additional
services could be easily and cheaply offered to the
Thin Client environment providing users with the
ability to access more compatible services from
within the Thin Client environment. Figure 4
identifies the key components within the design.
4.2.1 Design
The most important addition to the second case
study was the provision of additional services which
were similar to those available in PC labs. This was
to ensure that students could use this facility and
have an experience on a par with the PC labs. A new
domain was created where Unix and Windows
shared a common authentication process. Due to
difficulties integrating Unix and the existing
Windows authentication process, the new Domain
was built on the LDAP system with SAMBA
providing the link between the new Windows
Terminal Servers and the LDAP system. While
students could now use the same username and
password combination for Windows and Unix
systems this was not integrated into the existing
Windows authentication process. Students were still
required to have two sets of credentials, the first for
the existing PC labs, and the second for access to a
new domain containing a number of Windows
Terminal Servers and the original graphical Unix
desktop. While the Thin Clients now provided
Windows and Unix graphical desktops, the new
Windows Domain was also accessible from existing
PC labs via RDP connections to the Terminal
Servers. This allowed classes to be scheduled either
inside or outside of the Thin Client laboratory. In
addition to providing Windows Terminal Services
(WTS), student owned virtual machines were now
also available. Due to the fact that most services
were now available from all locations, the ease of
access to the services from within the Thin Client lab
was improved by providing users with a menu of
destinations upon login. This new login script
effectively provided a configurable redirection
service to the WTS and Virtualisation destinations
using the rdesktop utility [31] which performed a full
screen RDP connection to specified destinations. An
interesting outcome of this destination chooser was
that any RDP based destination could be included
regardless of the authentication process used. This
would however require a second authentication
process with the connecting service. The new
services provided were as follows:
a) A general purpose Windows Terminal Server
with mounted storage for all students and staff.
UbiCC Journal – Volume 4 No. 3
b) Course specific Windows Terminal Servers for
courses where there were specific software
requirements not common to all students.
c) Individual Virtualised desktops for students in
specific modules where administration rights
were required.
d) All services were made available from both the
Thin Client and PC labs as they were available
over the Remote Desktop Protocol RDP.
e) Provisioning of an easy access point to all
services from within the Thin Client environment
which was not available from PC systems.
4.2.2 Results
The data gathered for Case Study 2 was evaluated
under same three headings as per case study 1.
1) Login events on the Thin Clients
2) Reservation of the Thin Client facility.
3) The cost of maintaining the service.
25
20
Login Events per Day
15 Case Study 2
10
Case Study 1
5
0
08 Feb 22 Feb 08 Mar 22 Mar 05 Apr
Figure 5: User Login Event Comparison
Login events on the Thin Clients:
Figure 5 shows a comparison of activity during
the same time period for the two case studies. To
identify trends in the data a displacement forward
moving average was performed on the data as shown
in Eq. (1).
(1)
It is clear that for the same time period there was
a significant increase in the use of the system as the
number of login events increased by a factor of 4.
Once again the login events were extracted from the
Solaris server by parsing the output of the last
command.
Reservation of the Thin Client Facility:
The changes to the Thin Client facility were
announced at the start of the second academic
semester as a PC upgrade and the number of room
bookings increased as shown in Figure 6 from 6
hours a week to 20 hours a week. This was due to
the use of the room as a Windows based laboratory
592
Special Issue on ICIT 2009 Conference - Applied Computing
using the new WTS and virtualisation services.
8 process which was not well documented. For
Case Study 1
Hours per day
6 Case Study 2
4
2 though emails were sent to students and
0 were posted in the labs. Usernames were based
Mon Tue Wed Thurs Fri
Figure 6: Thin Client Room Reservations
The Cost of Maintaining the Service:
All of the benefits observed from the first case
study were retained within this case study. The
addition of terminal services reduced the reliance of
students on Fat Client installations. Students are now
using virtual machines and terminal servers on a
regular basis from all labs.
4.2.3 Analysis
This second case study certainly saw an
improvement over its earlier counterpart and students
and staff could now access more familiar services
from the Thin Client lab. Given the dramatic increase
relative to the earlier results it could be stated that
the introduction of the more familiar services
increased the acceptance of the facility. Both case
studies demonstrated equally well that it is possible
to obtain the total cost of ownership benefits using a
Thin Client model, but the services offered has a
dramatic affect on user acceptance. It is useful to
review the outcome in relation to the UTUAT.
a) Performance Expectancy
Given that new services such as personalised
virtual machines were now available, staff and
students could identify a clear advantage to the
system where administration rights could be
provided in a safe manner, allowing more
complex and previously unsupported activities
to take place. For example, the Advanced
Internet module for the MSc. students could now
build and administer full web servers which
could remain private to the student ensuring that
no other student could access or modify a
project which was a work in progress.
b) Effort Expectancy
Considerable improvements were made in this
case study to allow users to access well known
environments from both the Thin Clients and PC
systems. Students who were taught modules
using the new WTS or virtual environments were
trained on how to access the systems, and once
they used them they continued to do so
throughout the year. Those who did not have
UbiCC Journal – Volume 4 No. 3
modules being taught using these new services
were still required go through a new login/access
example within the Thin Client labs the new
username/password combination was required to
access the choice of destinations from the
devices. This acted as a barrier to use even
information on how to access these accounts
on existing student ID numbers.
c) Social Influence
Little changed in this case study for those who
did not have a teaching requirement based on the
new services.
d) Facilitating Conditions
With the provision of WTS services and virtual
machines which provided Windows
environments the issue of compatibility was
reduced. However two issues remained which
were not addressed. Firstly while users could now
share a common data store between systems on
this new domain there was no pre-packaged
access to the data store on the existing PC
domain. While it was technically possible to
combine both under a single view, this required
user intervention and additional training which
was not provided. Secondly the sequence of steps
required to access choices from the Thin Clients
was a non-standard login process which now
required a second login, the first of which was at
a Unix graphical login screen. For many this
initial login step remained as a barrier to using
the system.
The most striking result from this case study is
that while the second case study demonstrated
significant increase in acceptance and use, the PC
environments remained the system of choice for
students, as shown in Figure 7. In this graph we
show the typical use PC laboratory within the same
faculty. Thin Client use remained less than one third
of the use of the busiest computer laboratory. Thin
Clients are shown to be capable of providing services
equally well to both Windows and Unix users by
introducing the ability of students to access their own
private desktop from many locations, however this
feature alone was not enough to entice users from the
existing PC infrastructure. Clearly the introduction of
virtualisation to the infrastructure allowed new
services to be developed and used from Thin and Fat
clients which could be seen as a potential for
migrating users to a Thin Client/Virtualisation
model, which indeed is a future planned initiative.
The results show a definite increase in the use of the
Thin Client facilities with data being gathered from
the same period over both case studies to eliminate
any bias which might occur due to module schedule
593
Special Issue on ICIT 2009 Conference - Applied Computing

differences at different time periods during the year. locations. It was essential that the Thin Clients were
The timing and method used to announce the not to be identifiable by students if at all possible,
changes was critical to the increase in acceptance. and to co-locate them with existing PC systems. To
The announcement of the systems as a PC upgrade ensure that all devices behaved in a consistent
removed some of the barriers which existed for users manner to PCs they must boot and present the same
who did not feel comfortable with a Unix login screen as would be expected on a PC in the
environment but failed to attract a majority of the same location. To achieve this all Thin Client
students. devices with the exception of the SunRay systems
used a Preboot Execution Environment (PXE) [32]
80 boot process to connect to a Linux Terminal Server
70 Project server (LTSP). The server redirected the user
PC Lab 1
60 session to the correct WTS using rdesktop where the
user was presented with a Windows login screen
50 identical to those on adjacent PC systems.
Login Events per Day

40 The SunRay systems were run in Kiosk mode

30
20 Case Study 2
10
Case Study 1
0 however was not central to the design of the case
08 Feb 22 Feb 08 Mar 22 Mar 05 Apr
Figure 7: Comparison with PC Computer Labs
4.3 Case Study 3
The third case study was designed using the
experiences of the first two case studies and was
extended beyond the School of Computing. It was
aimed at demonstrating the capability of the Thin
Client technology in two different demographic
environments, the first was one of the Institute
Libraries where PCs were used by students from
many different faculties and the second was within
the Business faculty where computer system use was
provided in support of modules taught within that
faculty. This case study expressed the following
aims at the outset
1) To demonstrate the use of Thin Client technology
within the student population and determine the
level of student acceptance of that technology.
2) To implement a number of alternative
technologies in order to provide a point of
comparison with respect to their overall
performance and acceptance.
3) To determine the capability of the existing
network infrastructure to support Thin Clients.
which allowed the boot sequence to redirect the
session to a WTS also via the rdesktop utility. The
WTS were installed on a VMWare ESX Server to
allow rollback and recovery of the servers. This
study and only served as a convenience in sharing
hardware resources between multiple servers. The
only concern was the potential performance of the
WTS under a virtualised model. Given that the
applications were primarily productivity applications
such as word processing and browsing, and that the
maximum number of users allowable on any WTS
was 25 (based on the number of devices which were
directly connected to the WTS) this was considered
to be within the acceptable performance range of the
architecture. This assumption was tested prior to the
case study being made accessible to students with no
specific issues raised as to warrant further
restructuring of the architecture
Seventy five Thin Clients were deployed in six
locations. The following Thin Client devices were
used as shown in Figure 8 and Table 1.

4.3.1 Design
Unlike the previous case studies the aim was to
insert Thin Clients into the existing environment as
invisibly as possible. This meant that existing
authentication processes were to be maintained.
There were two different authentication processes in
place which needed to be support, Novell Client for
the Business faculty and Active Directory for the
Library. In both cases a WTS system was built which
joined to the respective domains. Applications were
installed on the Thin Client in order to mirror those
Figure 8: Case Study 3
that were present on existing PCs in the chosen

UbiCC Journal – Volume 4 No. 3 594

Special Issue on ICIT 2009 Conference - Applied Computing
Table 1: Thin Clients deployed
Device Boot Mode Quantity
Dell GX260 PXE Boot PC 15
Dell FX 160 PXE Boot TC 25
HP T5730 PXE Boot TC 8 the case study which was over one month during
Fujitsu FUTRO S PXE Boot TC 2 which data was collected from the three WTS
SunRay 270 SunRay 25
4.3.2 Linux Terminal Server Project
LTSP works by configuring PCs or suitable Thin
Clients to use PXE-Boot to obtain the necessary
kernel and RDP client used as part of this project.
These are obtained from a TFTP server whose IP
address is provided as a DHCP parameter when the
client PXE-Boots. As part of the DHCP dialogue,
devices configured to PXE-Boot are given settings
by the DHCP server. These include; TFTP Boot
Server Host Name and Bootfile Name.
The necessary settings were configured on each
of the DHCP servers serving the relevant locations
within the DIT so as to point any PXE-Boot devices
to the relevant LTSP boot server and to specify the
kernel to be loaded by the PXE-Boot client. Using
these settings the PXE-Boot clients load a Linux
kernel and then an RDP client which connects to one
of the three WTS used as part of this case study.
140
Login Events per Day
120
100
80 Library
Lab 2
60
40
Lab 1
20
0 75%
17 Apr 24 Apr 1 May 8 May 15 May
Figure 9: User Login Event Comparison
4.3.3 Results
Use of the Thin Clients was recorded using login
scripts on the Windows Terminal Servers which
recorded login and logout events. As expected the
use of the Library systems exceed the use of the
laboratories but both were in line with typical use
patterns expected for each location. What was
immediately obvious was that each location had a
higher utilization than the previous two case studies
but comparable with the PC labs shown in Figure 9.
One of the difficulties with the comparison however
is that the final case study was performed at a
different point in the teaching semester and use of
the systems declined as students prepared for
examinations. Lab 1 was a “quiet lab” located
UbiCC Journal – Volume 4 No. 3
remotely from the primary labs within the Business
faculty and traditionally did not have high use. Lab 2
was a more central location and again as expected
this exhibited greater user activity. The systems
remained in operation continually for the period of
systems.
4.3.4 User Survey
Once the case study was running a desktop
satisfaction survey which employed the Likert scale
[33] was conducted to obtain feedback from students
using the Thin Client systems. The design of the
questionnaire was such that students were asked to
identify their desktop using a colour coded system
which was known only to the authors. Each of the
Thin Clients and a selection of PC systems (which
were not PXE booted) where targeted for the survey
to allow a comparative analysis between all Thin
Clients and existing PC systems to be performed.
The survey did not reference Thin Clients in any of
the questions but rather sought feedback on
application use and overall satisfaction with the
performance of the system through a series of
questions. There were 234 responses recorded for the
survey. The key questions in the survey were as
follows.
1) Please rate the overall performance of the
machine you are currently using
2) Please identify the primary reason you used this
computer
3) How would you rate your overall satisfaction
with this desktop?
4) Would you use this desktop computer again?
80%
User Satisfaction Ratings
70%
65%
60%
PC-Fat SunRay PXE HP TC Dell TC
Client Boot PC
All Applications Browsers
Figure 10: User satisfaction rating of desktop performance
The issue of overall performance was broken
down by the device used which was identified using
the colour coded scheme described earlier. Figure 10
below represents the average rating of satisfaction
reported by users broken down by device and
primary application in use. Since over 50% of
responses identified “Browsing” as the primary
reason for using the machine there are two
595
Special Issue on ICIT 2009 Conference - Applied Computing
satisfaction ratings provided as a point of
comparison. Figure 11 shows the combined rating of
users responses to overall satisfaction with desktop,
desktop performance and application performance.
80%
User Satisfaction Ratings
78%
76%
74%
72%
70%
68%
66%
PC SunRay PXE
Boot
Figure 11: Combined rating of desktop performance
4.3.5 Analysis
This final case study while shorter in length than
the other case studies demonstrated significant
progress in user acceptance. As part of the survey
users were asked if they would consider reusing the
system and as can be seen in Figure 12 there was
significant support for the systems.
The small number of responses representing
those who did not wish to reuse the system cited
USB performance as the primary cause of their
dissatisfaction. This was identified early in the
testing of the Thin Clients that all systems performed
noticeably slower than the PC systems in this
respect. Questions regarding the primary storage
method used by students were added to the survey as
was a satisfaction rating. From the results in Figure
13 it is clear that while the PC systems did perform
better when users primarily used USB storage, the
satisfaction in storage performance for all other
options were comparable. The HP satisfaction rate
had a low survey response rate and hence was not
considered significant in our analysis given the small
number of data points.
NO, 8%
YES,
92%
Figure 12: User Response "Would you use this system
again"
UbiCC Journal – Volume 4 No. 3
Non-USB Storage USB Only
User Satisfaction Ratings
100%
90%
80%
70%
60%
50%
40%
PC SunRay PXE HP Dell
Boot
Figure 13: Storage Satisfaction Rating
By making the Thin Clients as invisible as
HP DELL possible and comparing satisfaction and user access
to the existing PC systems it was clear that for the
majority of users there was no apparent change to the
services provided. Integrating into the existing
authentication process was an essential feature of this
case study as was the presenting of a single
authentication process at the WTS login screen.
Efforts were also made to ensure that the
applications installed on the WTS were configured to
look and feel the same as those on the standard PC.
As with the previous case studies it is useful to
review the case study in relation to the UTUAT.
a) Performance Expectancy
With the exception of increasing the number of
desktops in the Library, the primary deployment
mainly replaced existing systems, so users were
not provided with any reminders that they were
using a different system. In effect there was no
new decision or evaluation by the user to address
the questions which were relevant in the previous
case studies.
b) Effort Expectancy
The reuse of the existing login/access procedure
which was well known and part of the normal
process for students using existing PC systems
again allowed for this factor to become mainly
irrelevant. Usernames, passwords, applications
and system behaviour were identical to those on
the PCs.
c) Social Influence
Without perceiving a difference in service, social
influence as a factor was also eliminated. Only
the SunRay systems had different keyboards and
screens, and as these screens were of higher
resolution than existing PCs they were if
anything seen as a more popular system.
d) Facilitating Conditions
Unlike the previous case studies support for the
facility was more complex. Different levels of
expertise and engagement were required. Thin
596
Special Issue on ICIT 2009 Conference - Applied Computing
Clients were now part of a larger support
structure where many individuals were not core
members of the technical team who built the
systems. However given that only three support
calls were raised during the case study there was
little pressure on this factor either. The calls
raised were not in fact directly related to the Thin
Client devices, but rather the network and the
virtual environments used to host the centralised
servers.
5 CRITICAL ANALYSIS
The UTUAT provides a useful reference point in
understanding some of the factors affecting
acceptance of the Thin Clients. In the first case study
the primary barrier to acceptance was the
incompatibility of the new system with the existing
system. Students were not motivated to use the new
system as there were few advantages to doing so and
considerable effort in learning how to use it. The
second case study while more successful still failed
to gain acceptance despite the expansion of services
offered being comparable with existing Windows
services. The session mobility and access from
anywhere feature, while useful did not overcome the
resistance of users to migrate to the Thin Clients.
Thin Clients still required separate credentials and
the login process was still different to the PC
systems. The third and final case study was designed
to provide the same existing services as the PC only
using a centralised server and Thin Client model. No
new services for the user were provided. The primary
aim was to have the systems indistinguishable from
the existing installation of PCs, effectively running a
blind test for user acceptance. Once the users
accepted the new systems, further machines could be
deployed quickly and cheaply. The total cost of
ownership and centralised support savings
demonstrated in the first two case studies were just
as relevant in the third case study.
6 CONCLUSION
There is considerable literature in support of Thin
Client technology, and while there may be debate
regarding the finer points of its advantages the issue
has been and continues to be one of acceptance.
Acceptance for Thin Clients as a technology is often
confused with the non technical issues arising from
the deployment. The UTUAT helps distinguish
between technical and non-technical issues and as
shown within our case studies, the way in which the
technology was presented to the user had a higher
impact on acceptance than had the technology itself.
This point is highlighted by the fact that the Thin
Client devices which were not widely used in first
case study were integrated seamlessly into the third
UbiCC Journal – Volume 4 No. 3
case study. These three case studies provide data
centric analysis of user acceptance and identify the
evolving designs of our deployments. To gain
acceptance of Thin Clients within an educational
institute our case studies identified these key factors.
1) Locate the Thin Clients among the existing PC
systems, do not separate them or isolate them.
2) Ensure that the login process and credentials
users are identical to the existing PC systems.
3) Ensure that the storage options are identical to the
existing PC systems
4) Focus on providing exactly the same services that
already exist as opposed to focusing on out new
services.
By ensuring we ran a blind test on the user
population where Thin Clients co-existed with PC
systems, and where the services offered were
indistinguishable by the user, we were able to show a
user satisfaction rating of 92%. No significant bias
was evident in our comparison of user attitudes of
desktop services delivered over PCs and Thin
Clients.
7 FUTURE WORK
Additional case studies are planned which will
focus on acceptance of Thin Clients within the
academic staff population and will evaluate the
relevance of some of the proposed core technological
advantages within that environment such as session
mobility, Desktop as a Service, and dynamic lab
reconfiguration and remote access using WAN and
not just LAN environments.
8 REFERENCES
[1] V. Venkatesh, M.G. Morris, G.B. Davis, and F.D.
Davis, “User acceptance of information
technology: Toward a unified view,” Mis
Quarterly, 2003, pp. 425-478.
[2] J.D. Northcutt, “CYB Newslog - Toward Virtual
Computing Environments.”
[3] D. Tynan, “Think thin,” InfoWorld, Jul. 2005.
[4] S.J. Yang, J. Nieh, M. Selsky, and N. Tiwari,
“The Performance of Remote Display
Mechanisms for Thin-Client Computing,” IN
PROCEEDINGS OF THE 2002 USENIX
ANNUAL TECHNICAL CONFERENCE,
2002.
[5] T. Richardson, F. Bennett, G. Mapp, and A.
Hopper, “Teleporting in an X window system
environment,” IEEE Personal
Communications Magazine, vol. 1, 1994, pp.
6-13.
[6] Citrix Systems, “Citrix MetaFrame 1.8
Backgrounder,” Jun. 1998.
597
Special Issue on ICIT 2009 Conference - Applied Computing
[7] Microsoft Corporation, “Remote Desktop
Protocol: Basic Connectivity and Graphics
Remoting Specification,” Technical White
Paper, Redmond, 2000.
[8] T. Richardson, Q. Stafford-Fraser, K. Wood, and
A. Hopper, “Virtual network computing,”
Internet Computing, IEEE, vol. 2, 1998, pp.
33-38.
[9] Microsoft Corporation, “Microsoft Windows NT
Server 4.0, Terminal Server Edition: An
Architectural Overview,” Jun. 1998.
[10] J. Nieh, S.J. Yang, and N. Novik, “A
comparison of thin-client computing
architectures,” Network Computing
Laboratory, Columbia University, Technical
Report CUCS-022-00, 2000.
[11] R.A. Baratto, L.N. Kim, and J. Nieh, “Thinc: A
virtual display architecture for thin-client
computing,” Proceedings of the twentieth
ACM symposium on Operating systems
principles, ACM New York, NY, USA, 2005,
pp. 277-290.
[12] B.K. Schmidt, M.S. Lam, and J.D. Northcutt,
“The interactive performance of SLIM: a
stateless, thin-client architecture,” Proceedings
of the seventeenth ACM symposium on
Operating systems principles, Charleston,
South Carolina, United States: ACM, 1999,
pp. 32-47.
[13] M. Annamalai, A. Birrell, D. Fetterly, and T.
Wobber, Implementing Portable Desktops: A
New Option and Comparisons, Microsoft
Corporation, 2006.
[14] “MokaFive, Virtual Desktops,”
http://www.mokafive.com/.
[15] R. Chandra, N. Zeldovich, C. Sapuntzakis, and
M.S. Lam, “The Collective: A cache-based
system management architecture,”
Proceedings of the 2nd USENIX Symposium
on Networked Systems Design and
Implementation (NSDI’05).
[16] M. Weiser, “How computers will be used
differently in the next twenty years,” Security
and Privacy, 1999. Proceedings of the 1999
IEEE Symposium on, 1999, pp. 234-235.
[17] M. Jern, “"Thin" vs. "fat" visualization clients,”
Proceedings of the working conference on
Advanced visual interfaces, L'Aquila, Italy:
ACM, 1998, pp. 270-273.
[18] S. Kissler and O. Hoyt, “Using thin client
technology to reduce complexity and cost,”
Proceedings of the 33rd annual ACM
SIGUCCS conference on User services, ACM
New York, NY, USA, 2005, pp. 138-140.
[19] M. Jern, “"Thin" vs. "fat" visualization clients,”
L'Aquila, Italy: ACM, 1998, pp. 270-273.
[20] S. Kissler and O. Hoyt, “Using thin client
technology to reduce complexity and cost,”
New York, NY, USA: ACM, 2005, pp. 138–
140.
UbiCC Journal – Volume 4 No. 3
[21] J. Golick, “Network computing in the new thin-
client age,” netWorker, vol. 3, 1999, pp. 30-
40.
[22] S.C. Speer and D. Angelucci, “Extending the
Reach of the Thin Client.,” Computers in
Libraries, vol. 21, 2001, pp. 46 - .
[23] P.A. Strassmann, “5 SECURE REASONS FOR
THIN CLIENTS.,” Baseline, 2008, p. 27.
[24] G.A. Plan, “An inefficient truth,” PC World,
2007.
[25] M. Bruno-Britz, “Bank Sheds Pounds.,” Bank
Systems & Technology, vol. 42, 2005, p. 39.
[26] “Sun Ray White Papers,”
http://www.sun.com/sunray/whitepapers.xml.
[27] B.K. Schmidt, M.S. Lam, and J.D. Northcutt,
“The interactive performance of SLIM: a
stateless, thin-client architecture,” Charleston,
South Carolina, United States: ACM, 1999,
pp. 32-47.
[28] S. Potter and J. Nieh, “Reducing downtime due
to system maintenance and upgrades,” San
Diego, CA: USENIX Association, 2005, pp. 6-
6.
[29] I. Wong-Bushby, R. Egan, and C. Isaacson, “A
Case Study in SOA and Re-architecture at
Company ABC,” 2006, p. 179b.
[30] G. Reynolds and M. Gleeson, “Towards the
Deployment of Flexible and Efficient Learning
Tools: The Thin Client,” The Proceedings of
the 4th China-Europe International
Symposium on Software. China (Guanzhou).
Sun Yat-Sen University. (2008).
[31] “rdesktop: A Remote Desktop Protocol client.”
[32] B. Childers, “PXE: not just for server networks
anymore!,” Linux J., vol. 2009, 2009, p. 1.
[33] R. Likert, “A Technique for the Measurement of
Attitudes,” Archives of Psychology, vol. 140,
1932, pp. 1–55.
598
Special Issue on ICIT 2009 Conference - Applied Computing

AN INTERACTIVE COMPOSITION OF
WORKFLOW APPLICATIONS BASED ON UML
ACTIVITY DIAGRAM

Yousra Bendaly Hlaoui, Leila Jemni Ben Ayed

Research Unit of Technologies of Information and Communication
Tunis, Tunisia
Yousra.bendalyhlaoui@esstt.rnu.tn
Leila.jemni@fsgt.rnu.tn

ABSTRACT
In today's distributed applications, semi automatic and semantic composition of
workflows from Grid services is becoming an important challenge. We focus in
this paper on how to model and compose interactively workflow applications from
Grid services without considering lower level description of the Grid environment.
To reach this objective, we propose a Model-Driven Approach for developing such
applications based on semantic and syntactic descriptions of services available on
the Grid and abstract description provided by UML activity diagram language as
well. As there are particular needs for modeling composed workflows interactively
from Grid services, we propose to extend the UML activity diagram notation.
These extensions deal with additional information allowing an interactive and semi
automatic composition of workflows. In addition this specific domain language
contains appropriate data to describe matched Grid services that are useful for the
execution of the obtained workflows.

Keywords: Grid services, Interactive, semantic, composition, Workflow

application, UML activity diagrams.

1 INTRODUCTION application and on the other reduces the complexity

Today’s distributed applications [23] are
developed by integrating web or Grid services [13,
14] in a workflow. Due to the very large number of
available services and the existence of different
possibilities for constructing workflow from
matching services, the problem of building such
applications is usually a non trivial task for a
developer. This problem requires finding and
orchestrating appropriate Grid services in a
workflow. Therefore, we propose an approach that
allows semi automatic and interactive composition of
workflow applications from Grid services. To
describe and model workflow applications we use
UML [25] activity diagrams. Recently, several
solutions were proposed to compose applications
from Grid services such as works presented in [8, 17,
18]. However, the proposed solutions need
interaction with user and guidelines or rules in the
design of the composed applications. Consequently,
the resulting source code is neither re-usable nor it
promotes dynamic adaptation facilities as it should.
However, for applications composed of Grid services,
we need an abstract view not only of the offered
services but also of the resulting application [31].
This abstraction allows the reuse of the elaborated
of the composed applications. There are several
architectural approaches for distributed computing
applications [22] which make easy the development
process. However, these approaches need rigorous
development methods to promote the reuse of
components in future Grid development application
[16]. It has been proven from past experience that
using structured engineering methods makes easy the
development process of any computing system and
reduces the complexity when building large Grid
application [22].
To reduce this complexity and allow the reuse of
Grid service applications, we adopt a model-driven
approach [24]. Thus we introduce in this paper a new
approach to build, interactively, workflow
applications by following OMG(s) principals of the
MDA in the development process [2, 3, 4].
In this approach [2, 3, 4], our focus is to
compose and model workflows from existing Grid
services that represent the main aspect in the
development of Grid services applications. The
workflow modeling identifies the control and data
flows from one depicted Grid service's operation to
the next to build and compose the whole application.
To model and express the composed workflow of
Grid services, we use as abstract language the

UbiCC Journal – Volume 4 No. 3 599

Special Issue on ICIT 2009 Conference - Applied Computing
activity diagrams of UML [25]. The provided model
forms the Platform Independent Model (PIM) of the
proposed MDA approach. This model is more
understandable for the user than an XML [35] based
workflow description languages like BPEL4WS [15]
which represent the Platform Specific Model (PSM).
This paper is organized as follows. Section 2
presents the related work. Section 3 introduces the
different components of the composition system;
section 4 specifies our proposed UML profile,
composition patterns and different steps of the
interactive composition process. Finally, section 5
concludes the paper and proposes areas for further
research.
2 RELATED WORK
Many works were carried out in the field of
Grid and Web services composition, such as works
presented in [8, 17, 18, 19, 20, 28, 29, 30]. In [28]
authors were interested in the semi automatic
composition of web services and proposed a
validation approach based on the semantic
descriptions of services and on a logic based
language to describe and validate the resulting
composite Web services. However, the resulting
composed web service is not clear for user who is
not familiar with logic based languages. In our
contribution, we propose a solution not only to
compose workflows from available Grid services,
but also to provide graphical and comprehensive
models of the resulting workflows. In the same
framework, authors in [29] proposed a composition
approach of Web services based on Symbolic
Transition Systems (STS). They developed a sound
and complete logical approach for identifying the
existence of available composition. They have
emphasized upon the abstract representation of the
composition request (the goal of the composition)
and the representation of the resulting composite
Web service. For the representation, authors have
used UML state machine diagrams [25] which are
suitable only to describe a sequence of component
services without addressing the other forms of
matching services in a workflow such as parallel
branches or and-branches. On the other hand, UML
activity diagrams that we use in our modelling
approach support all kind of workflow composition
patterns [10] such as parallelism, split and fork. The
authors in [19, 20, 30] have proposed a Model
Driven Approach for composing manually Web
services. They were based on UML activity diagrams
to describe the composite Web service and on UML
class diagrams to describe each available Web
Service. The user depicts the suitable Web service
and matches it in the workflow representing the
composite Web service using UML activity
diagrams. This approach would have been better if
the composition were automatically elaborated, since
UbiCC Journal – Volume 4 No. 3
the number of available services is in increase with
the existence of several forms and manners to
compose such services.
Based on domain ontology description, we lead
the user through to the composition process. Also,
we provide for this user a graphical interface based
on a domain specific UML language for automatic
grid service composition. This UML profile [5] is
based on stereotypes, tagged values and workflow
patterns [5] that we propose to ensure the automatic
composition. In the field of Grid services
composition the most related work is the work
presented by Gubala et al in [8, 17, 18]. In this work,
the authors have developed a tool for semi automatic
and assisted composition of scientific Grid
application workflows. The tool uses domain specific
knowledge and employs several levels of workflow
abstractness in order to provide a comprehensive
representation of the workflow for the user and to
lead him in the process of possible solution
construction, dynamic refinement and execution. The
originality of our contribution is that firstly we save
the effort of the user from the dynamic refinement
and execution as we propose a Model Driven
Approach which separates the specific model from
the independent model.
Secondly, we use UML activity diagrams to
deliver the functionality in a more natural way for
the human user. The use of UML activity diagrams
in the description of workflow application is argued
in several works such as works presented in [1, 10,
12, 27]. Thus, the advantage of UML activity
diagrams is that they provide an effective visual
notation and facilitate the analysis of workflows
composition.
In our approach, we propose an UML profile for
composing systematically a workflow application
from Grid services [5].
3 THE INTERACTIVE WORKFLOW
COMPOSITION SYSTEM
The system allows an interactive and semantic
composition of workflows from Grid services. As
shown in figure 1, the system is composed of three
components: a Grid Services workflows composer,
an ontological Grid Services registry and a
workflows execution system also we call it activity
machine.
3.1 The Grid services workflow composer
This system is composed of three components:
the composition tool, the transformation tool and the
verification tool.
3.1.1 The composition tool
It provides a graphical interface in the form of
UML activity diagrams editor allowing to the user an
interactive, systematic and semantic workflow
600
Special Issue on ICIT 2009 Conference - Applied Computing
Figure 1: Different components of the workflow composition system
Composition [6]. This composition is based on the
composition process which will be detailed in
section 4.3. In the Grid registry, services are
described in an ontological form with statements
regarding the service operation's inputs, outputs,
pre-conditions and effects (the IOPE set) [26].
Through these notions, the composition system is
able to match different grid service’s operations
into a workflow following a reverse traversal
approach. Thus, and by associating the required
data with the produced output, the composer
constructs a data flow between Grid service’s
operations using our workflow composition
patterns and UML profile[5]. The composer may
also use a specific notion of effect that may bind
two operations together with non-data dependency.
If the Grid registry fails to find the right operation,
the composition process stops. Otherwise, the
composition process will stop when all workflow
dependencies are resolved.
The request is sent to the Ontological Grid
Registry in the form of SPARQL query [34]. This
language provides a higher-level access to the
ontology transcribed knowledge for the automatic
discovery and semantic matching of services.
Therefore, once the workflow model is built, it
should be validated and verified to ensure its
reliability before being executed and reused as sub-
workflow.
3.1.2 The transformation tool
To support the verification and the execution of
workflow models described in UML activity
diagrams (UML-AD), the transformation tool
UbiCC Journal – Volume 4 No. 3
translates the activity diagram into a Hyper-Graph
(HG). This HG will be translated as well by the
transformation tool into a NuSMV format file
according to a relative semantic. The details of
these semantics may not be relevant to the topic for
which the paper is submitted. However these details
could be made available.
3.1.3 The verification tool
Checking errors in design models like UML
activity diagrams is essential since correcting an
error while the system is alive and running is
usually very costly [21]. Consequently, workflow
activity diagram models should be spotted and
corrected as early as possible [6].
Several techniques are used in the field of
behavioural design verification such as theorem
proving and model checking [11]. The latter is the
most useful because it is fully automatic and gives
feedback in case of detected errors. It verifies
whether some given finite state machine satisfies
some given property specified in temporal logic [9].
For activity diagrams, symbolic model checking
has proven to be an efficient verification technique
[11]. Thus, our verification tool is based on
NuSMV symbolic model checker [9] that supports
strong fairness property which is necessary to be
verified in a workflow model to obtain realistic
results. With the model checker, arbitrary
propositional requirements can be checked against
the input model. If a requirement fails to hold, an
error trace is returned by the model checker. The
transformation tool translates systematically the
error trace into an activity diagram trace by high-
601
Special Issue on ICIT 2009 Conference - Applied Computing
lighting a corresponding path in the given activity
diagram.
3.2 The Grid registry
During the workflow composition process, the
Grid registry provides the composer system with
the description of services available at the moment
and provides reasoning capabilities to enable proper
matchmaking of services inputs and outputs.
The Grid registry [6] is an ontological
distributed repository of Grid services and sub-
workflows. This registry is responsible for storing
and managing documents which contains
descriptions of syntax and semantics of services
and their operations expressed in an RDF file [33].
The semantic Web is making available technologies
which support automate knowledge sharing. In
particular there are several existing initiatives such
as OWL-S [26] which proves that ontologies have a
key role in the automating service discovery and
composition. That knowledge is based on semantic
descriptions of service classes published by the
service developers and provided in the Grid
environment [16]. Our Grid registry is based on an
ontological description of services and workflows.
The service ontology [7] provides concepts
and properties that allow description and
matchmaking of available services. A part of this
ontology is common to all services and it is based
on a standard semantic web service description
ontology OWL-S [26] which makes interoperability
with existing services. A part from the common
ontology, there is a domain specific part of the
ontology. The domain service ontology [7] allows
users to extend the common ontology schema in
order to provide a better specification of services as
well as their inputs and outputs. For these we define
a data ontology [7] which provides concepts and
properties for describing services input and outputs.
Ontology alignment [7] is a process for finding
semantic relationships among the entities of
ontologies. Its main activity is to find similar
concept in ontologies being aligned, in order to map
them. The measures for similarity computation can
be divided into two general groups; namely lexical
measures and structural measures. Lexical measures
are based on surface similarity such as title or label
of entities. In contrary, structural measures try to
recognize similarities by considering structures of
ontology graphs. The most advanced similarity
algorithms use combination of multiple similarity
measures to obtain more information about
concepts similarity. In our Grid registry, we adopt
an approach using a combination of lexical and
structural similarity [7].
We use similarity measures for mapping domain
ontology as initial selection and then the selection
will be refined with using structural similarity
method [7].
UbiCC Journal – Volume 4 No. 3
3.3 The workflow execution system
The reliable workflow model is sent to the
workflow execution system [6] which produces
implementation code for handling control flow and
data flow. The activity diagram describing the
workflow model is translated into a specific XML
file which will be the input of the execution system.
A workflow execution system executes different
workflow activities specified in the workflow XML
document in the correct order and with their
required inputs and outputs data. The execution of
an activity corresponds to the invocation of a Grid
service’s operation. The workflow execution
system monitors these activities using the tagged
values information expressed in the activities but
does not perform them. An activity of the activity
diagram modelling the workflow represents a state
of the workflow execution system in which the
system waits for an invoked grid service operation
to complete its work. Hence, the defined semantics
of activity diagrams for the verification describe the
behaviour of the execution system. When the
system enters a state relative to an invocation grid
service node or activity ai, it invokes a piece of
behaviour that is executed by the service or system
environment. While the latter is in ai (activity ai is
active), it waits for the termination event of the
invoked piece of behaviour. When termination
event occurs, the system reacts by executing the
outgoing edge E: it leaves the E's sources and enters
the E's targets and the execution process continues
for the other activity nodes until the final node is
reached.
4 UML BASED INTERACTIVE
COMPOSITION OF WORKFLOWS
FROM GRID SERVICES
In order to match and compose different Grid
service’s operations, we need to analyze constructs
of workflow models at higher abstraction level.
Since UML [25] is the core of the MDA [24], we
use its activity diagram language to model
composed workflows. The composition system
provides to the user a graphical interface to
compose its request using a UML profile specific
for the domain of composing systematically
4.1 UML Profile for composing workflows
In this section, we present our UML profile
which is based on Domain Specific Language
(DSL) for customizing UML activity diagrams for
the systematic composition of workflows from Grid
services [5].
In our DSL (See Figure 2), an activity of an
UML activity diagram represents a Grid service's
operation, while object flows represent the types of
results which flow from one activity to another.
602
Special Issue on ICIT 2009 Conference - Applied Computing
Effects binding two operations are presented with
control flows [5].
The name of an activity in the diagram
represents the name of the Grid service's operation.
This name must be specified as a Grid service could
have more than one operation often called interface
which are specified in its relative WSDL file [32].
There are two different types of activities: yet-
unresolved activities and established activities of
the composed workflow. The former represent the
need for a Grid Service's operation to be inserted in
order to complete the workflow. However, the latter
represent abstract operations that are already
included into the workflow.
As there are two different activity types in a
Grid service workflow model, an activity needs to
be typed and specified. To fulfil this, we propose to
use the DSL modelling element invoke to
stereotype an established activity which is used to
invoke an external Grid service's operation and yet-
unresolved to stereotype activities which are not yet
resolved. Object nodes of an established activity are
data stereotyped. Unknown input and output for a
yet-unresolved activity are unknown stereotyped.
In our UML profile, an object node could be
relatedto a final node as composed workflow of
Grid application should always deliver a result.
4.2 UML-AD composition patterns
We identify, in this section, how UML activity
diagrams support some of basic Grid service
composition patterns [5]. These patterns are
Figure 2: Meta-model of Grid service workflow composition specific UML activity diagram language
UbiCC Journal – Volume 4 No. 3
essential in the systematic building of workflow
applications from Grid services. The use of these
patterns depends on the number of the depicted
Grid Service's operations and their inputs and
outputs [5]. These operations are results of the
semantic research elaborated by the ontological
Grid services registry. This research is invoked by a
request given by the composition system in order to
complete an unresolved activity in the workflow.
The Grid service registry provides zero, one or
more operations producing the intended output.
Operations are depicted to be inserted in the
workflow interactively with the user.
4.2.1 Sequence Pattern
When the Grid registry provides one Grid
service's operation that is able to produce the
required result or the user selects one operation
from the provided operation set; the composition
system uses the sequence pattern to insert the
operation in the workflow. In this case and as is
illustrated by the figure 3, a single abstract
operation or activity (e.g. GridService1Operation1)
will be inserted in the workflow model described by
the UML-AD language. This operation may also
require some data for itself (e.g.
GridService1Operation1Input) and thus it may
introduce a new unresolved dependency (e.g. the
yet-unresolved stereotyped activity). So, we use a
follow-up method to build a simple pipeline-type
sequential workflow: a sequence pattern.
603
Special Issue on ICIT 2009 Conference - Applied Computing
A Sequence pattern is composed with sequential
activities which are related with control flow (non
data operations dependency) or object flow (data
operation dependency).
Figure 3: The sequence pattern
4.2.2 And-branches pattern
The and-branches pattern is introduced when the
introduced operation represented by an abstract
UML activity has more than one input. This pattern
is based on the Synchronization pattern presented in
[9].
This pattern starts with object nodes,
representing alternative operation inputs, which flow
to a join node. The latter is linked to the abstract grid
service's operation. This operation introduces some
unresolved dependencies in the workflow.
Semantically, several services instances are
invoked in parallel threads and the join will wait for
all flows to finish. As illustrated in Figure 4, the
operation of the Grid service
GridService1Operation1 needs two inputs data
GridService1Operation1Input1 and
GridService1Operation1Input2. The relative pattern
produces two parallel threads in the workflow.
Figure 4: And-branches pattern
4.2.3 Alternative branches pattern
When the Grid registry provides more than
one operation able to produce the required result, and
the user do not select one of them, the composition
requires a specific pattern: the alternative branches
pattern.
This pattern combines the Exclusive Choice
and Simple Merge patterns presented in [9]. In this
pattern, each alternative service's operation is linked
UbiCC Journal – Volume 4 No. 3
to object node representing the required input which
both of them flow to a merge construct.
Semantically, several services instances are
invoked in parallel threads and the merge will only
wait for the first flow to finish. We distinguish, in
Figure 5, two different Grid service's operations,
GridService1Operation1 and GridService2Operation1
providing the same output data DataOutput.
Figure 5: Alternative branches pattern
4.2.4 Alternative services pattern
When composing workflows from Grid
services, a specific matching based on semantic
comparison could provide two or more different Grid
services performing each of them the required
operation. In such case and when the user do not
choose one of the depicted Grid service’s operations,
the composition system uses the alternative services
pattern to involve the operations in the workflow
model.
In this pattern, the Grid service’s operation to
insert is modelled by a composed super-activity with
a specified input data object and specified output
data object (Figure 6). The super-activity is
stereotyped as AlternativeServiceInstance to indicate
that its task may be accomplished by a set of
alternative service's instances. These alternative
service instances are described with sub-activities.
The sub-activities shall be grid service instances and
thus stereotyped as invoke. It was up to decision
mechanism of the workflow execution engine to
choose which service instance in such given
workflow node is to be invoked and executed. In
Figure 6, the data DataOutput is provided from
GridServiceOperation service operation which could
be GridService1Operation1 provider or
GridService2Operation2 provider.
Figure 6: Alternative services pattern
604
Special Issue on ICIT 2009 Conference - Applied Computing
4.3 The composition process
Figure 7 illustrates the scenario of the
composition process of workflows from available
Grid services. This composition is based on the
domain specific UML activity diagram language
presented in section 4.1. In the following, we
comment the different process steps of the scenario
presented in the figure 7.
Step 1: The user builds its composition request by
specifying what kind of outcome or result that it
expects from the workflow application execution.
Step 2: The composition system analyses the desired
output and sends a SPARQL query to request the
ontologies of the Grid registry describing the
available Grid services. The composer requests the
Grid registry for a Grid service’s operation having
the specified result as output.
Step 3: If the required operation is not found and all
unknown results are resolved then the composition
process stops.
Figure 7: Scenario of the interactive composition of Grid service workflows based on UML activity diagrams
UbiCC Journal – Volume 4 No. 3
Step 4: If the required operation is found then the
system displays its characteristics to the user to
confirm the choice. The register may provide more
than one operation. In such case the user could
choose the operation to insert in the workflow model
from the given list. If it does not specify its
operation, then the system inserts all the given
operations using one of the composition patterns
presented in section 4.2. Relatively to the number of
depicted operations and their inputs and outputs, the
composer chooses the right composition pattern.
Step 5: For each input of inserted operation, the
system defines one unresolved dependency as a
workflow activity which is not yet established. This
activity depends on some Grid service’s operation.
For each unresolved dependency the composer asks
the user if it wants to continue the composition
process or not. If the response is positive the
composer re-executes the process from the step 2 to
resolve the current unresolved dependency.
605
Special Issue on ICIT 2009 Conference - Applied Computing
5 Illustration of the interactive workflow
composition
In the following, we illustrate the composition
process through the example of the domain of the
city traffic pollution analysis. This application, as
presented in [23], targets the computation of traffic
air pollutant emission in an urban area.
Step 1: Figure 8 shows an example of initial
workflow that represents a composition request for
the results of the pollutant emission due to the city
traffic. The desired result, PollutionEmission, is
described by the rectangle representing the object
node in the relative activity diagram.
Figure 8: Initial workflow as a composition request
Step 2: Figure 9 represents the workflow of the
computation of traffic air pollution analysis after one
step of composition. The service’s operation,
delivering the PollutionEmission result, is
AirPollutionEmissionCalculator. This operation is
the result of the composer query asked to the
ontological Grid registry. The operation requires two
inputs TrafficFlowFile and PathsLenght-File, thus
it infers two unresolved dependencies in the activity
diagram modelling the composed workflow.
Figure 9: An example of workflow after one step of
composition
Step 3: For every dependency that needs to be
resolved .i.e. a yet-unresolved activity, the composer
contacts the ontological registry in order to find
suitable service’s operations that may produce the
required result. The services are described in an
ontological form with statements regarding the
service operation’s inputs, outputs, preconditions and
effects (the IOPE set) [26]. Through these notions,
UbiCC Journal – Volume 4 No. 3
the composer system is able to match different
operations into a workflow following a reverse
traversal approach. Thus, and by associating the
required data with the produced output, the composer
constructs a data flow between operation using
workflow patterns and our UML profile [5]. The
composer may also use a specific notion of effect
that may bind two operations together with non-data
dependency. In [10], five basic control patterns were
defined to be supported by all workflow languages
and workflow products. These patterns are Sequence
pattern, Parallel split pattern, Synchronization
pattern, Exclusive Choice and Simple Merge
patterns. Figure 10 represents the example of city
traffic analysis Workflow after the full composition
activity. It involves several Grid service operations,
sequence branches, parallel split branches, simple
merge branches and a loop [5]. The loop is involved
in the workflow diagram as the application iterates in
order to analyze the possible traffic. The Figure
shows also how UML activity diagrams support the
five basic patterns in the composition specific
domain of Grid services workflows [5]. In the
example, some of object node or input data, such as
VehiculeType and StartZonzId, are given by the user
of the application; they do not have an operation
provider. This illustrates the interaction between our
composition system and the user.
Figure 10: The workflow application after the full
composition
606
Special Issue on ICIT 2009 Conference - Applied Computing
6 CONCLUSION
In this paper, we have presented an approach for
composing interactively workflows from Grid
services [2, 3, 4, 6]. This composition is based on an
UML profile for customizing UML activity diagrams
to compose and model workflows [5] and on
composition patterns [5] as well. The interactive
composition process was illustrated through the
example of city traffic pollution analysis domain [23]
We have developed and implemented most of the
presented components of the composition system.
Actually, we are working on the implementation
of the workflow execution system that invokes and
executes the depicted Grid service instances and
manages the control and data flows in a run time
environment relatively to our proposed activity
diagram semantic.
7 REFERENCES
[1] R. Bastos, D. Dubugras, A. Ruiz: Extending
UML Activity Diagram for Workflow modelling
in Productions Systems, In. Proc. Of the 35th
Annual Hawaii International Conference on
System Sciences, HICSS'02, IEEE Cs Press
(2002).
[2] Y. Bendaly Hlaoui, L. Jemni Ben Ayed: Toward
an UML-based composition of grid services
workflows, In Proc. Of the 2nd international
workshop on Agent-oriented software
engineering challenges for Ubiquitous and
Pervasive Computing, AUPC’08, ACM Digital
Library, pp. 21-28 (2008).
[3] Y. Bendaly Hlaoui, L. Jemni Ben Ayed: An
extented UML activity Diagram for Composing
Grid Services Workflows, In Proc. Of the IEEE
international Conference on Risks and Security
of Internet and Systems, CriSIS’08, Tozeur,
Tunisia, p. 207-212 (2008).
[4] Y. Bendaly Hlaoui, L. Jemni Ben Ayed: A
MDA approach for semi automatic grid services
workflows composition, In Proc. Of the IEEE
international conference on Industrial
Engineering and engineering Managment,
IEEM’08, p.1433-1437 (2008).
[5] Y. Bendaly Hlaoui, L. Jemni Ben Ayed: Patterns
for Modeling and Composing Workflows from
Grid Services, In. Proc. Of the 11th International
Conference on Enterprise Information Systems,
ICEIS'2009, Milan, Italy, LNBIP, Springer-
Verlag, Vol. 24, pp. 615-626 (2009).
[6] Y. Bendaly Hlaoui, L. Jemni Ben Ayed: An
UbiCC Journal – Volume 4 No. 3
Interactive Composition of UML-AD for the
Modelling of Workflow Applications, In. Proc.
Of the 4th International Conference on
Information Technology, ICIT'2009, Amman,
Jordan (2009).
[7] Y. Bendaly Hlaoui, L. Jemni Ben Ayed:
Ontological Description of Grid Services
Supporting Automatic Workflow Composition,
In. Proc. Of the International Conference on
Web and Information Technologies,
ICWIT'2009, Kerkennah, Tunisia, ACM
SIGAPP.fr, IHE éditions, pp. 233-243 (2009).
[8] M. Bubak, R. Guballa, M. Kapalka, M.
Malawski, K. Rycerz: Workflow Composer and
service registry for grid applications, Journal of
Future Generation Computer Systems, Vol. 21,
pp. 79-86 (2005).
[9] A. Cimatti, E. Clarck, A. Tacchella: Nusmv
version 2: An opensource tool for symbolic
model checking, In Proc. Of the International
Conference on Computer-Aided Verification,
CAV'02, Lecture Notes in Computer Science,
Springer Verlag (2002).
[10] M. Dumas, and A. H. M. ter Hofsetde: UML
Activity Diagrams as a Workflow Speci_cation
Language, In UML'2001 Conference, Toronto,
Ontario, Canada, Lecture Notes in Computer
Science (LNCS), Springer-Verlag, Heidelberg,
Germany (2001).
[11] R. Eshuis: Semantics and verification of UML
Activity Diagrams for Workflows Modelling,
PhD thesis, University of Twente (2002).
[12] R. Eshuis and R. Wieringa: Comparing Petri net
and Activity diagram variants for workflow
modelling: A Quest for Reactive Petri Nets,
Petri Net technology for communication based
Systems, LNCS, Springer Verlag (2003).
[13] I. Foster, D.Berry, A.Djaoui, A.Grimshaw,
B.Horn, H.Kishimoto, F.Maciel, A.Savy,
F.Siebenlist, R.Subramaniam, J.Treadwell,
J.Von Reich: The Open Grid Services
Architecture, Version 1.0. (2004).
[14] I. Foster, C. Kesselman: Grid Services for
Distributed System Integration, Journal of IEEE
Computer, Vol. 35, No. 6, pp. 37-46 (2004).
[15] T. Gardner: UML modelling of automated
Business Processes with a Mapping to
BPEL4WS, In. Proc. Of the European
Conference on Object Oriented Programming
(2003).
607
Special Issue on ICIT 2009 Conference - Applied Computing
[16] C. Goble, D. de Roure: The grid: an application
of the semantic web, ACM SIGMOD Record-
Special section on semantic web and data
management, Vol. 31, No.4, pp. 65-70 (2002).
[17] T. Gubala, D. Herezlak, M. Bubak, M.
Malawski: Semantic Composition of Scientific
Workflows Based on the Petri Nets Formalism,
In Proc. Of the Second IEEE International
Conference on e-Science and Grid Computing,
e-Science'06 (2006).
[18] R. Guballa, A. Hoheisel, F. First: Highly
Dynamic workflow Orchestration for scientific
Applications, CoreGRID Technical Report,
Number TR-0101 (2007).
[19] R. Gronomo, I. Solheim: Towards Modelling
Web Service Composition in UML, In The 2nd
International Workshop on Web Services:
Modelling, Architecture and Infrastructure, Porto,
Portugal (2004).
[20] R. Gronomo, MC. Jaeger: Model Driven
Semantic Web Service Composition, In Proc. Of
the 12th Asia-Pacific Software Engineering
Conference, APSEC'05 (2005).
[21] M. Laclavik, E.Gatial, Z. Balogh, O. Habala, G.
Nguyen, L.Hluchy: Experience Management
Based on Text Notes, In. Proc. Of e-Challenges
2005 Conf. (2005).
[22] W. Li, C. Huang, Q. Chen, H. Bian: A Model-
Driven Aspect Framework for Grid Service
Development, In Proc. Of the IEEE International
Conference on Internet and Web Applications
and Services, ICIW’06, pp. 139-146 (2006).
[23] M. Masetti, S. Bianchi, G. Viano: Application of
K-Wf Grid technology to Coordinated Traffic
Management. http://grid02.softeco.it/site/project-
info.html
[24] Model Driven Architecture (MDA). Document
nomber omrsc/2001-07-01 (2001)
[25] Object Management Group. UML 2.0
Superstructure Specification. July (2005).
[26] OWL-S: Semantic Markup for Web Services.
The OWL Services Coalition. OWL-S version
2.0.S.
UbiCC Journal – Volume 4 No. 3
[27] Pllana, T. Fahringer, J. Testori, S. Benkner, I.
Brandic: Towards an UML Based Graphical
Representation of Grid Workflow Application,
In. Proc. Of the 2nd Eu-ropean Across Grids
Conference, Nicosia, Cyprus, Springer-Verlag
(2004).
[28] J. Rao, P. Kungas, M. Matskin: Logic-based
web service composition: from service
description to process model, In Proc. Of the
IEEE International Conference on Web Services,
ICWS 2004, San Diego, California, USA (2004).
[29] E. Sirin, J. Hendler, B. Parsia: Semi automatic
composition of web services using semantic
descriptions, In Proc. Of the ICEIS-2003
Workshop on Web Services, Modeling,
Architecture and Infrastructure, Angers, France
(2003).
[30] D. Skogan, R. Gronomo, I. Solheim: Web
Service Composition in UML, In Proc. Of the
8th Intl Enterprise Distributed Object Computing
Conference, EDOC'04 (2004).
[31] M. Smith, T. Friese, B. Freisleben: Model
Driven Development of Service-Oriented Grid
Applications, In. Proc. Of the IEEE Asia-Pacific
Conference on Services Computing, APSCC'06
(2006).
[32] Web Services Descriptio Language (WSDL)
1.1. W3C Note 15 March (2001).
[33]W3C: Resource Description Framework (RDF)
Model and Syntax Specification, report num.
TR/1999/REC-rdf-syntax-19990222 (1999).
[34]W3C: SPARQL Query Language for RDF,
report , 2008.
[35] M. J. Young: XML Step by Step, Microsoft
Press, ISBN: 2-84082-812-X (2001).
608
Special Issue on ICIT 2009 Conference - Applied Computing

HOW TO MAP PERSPECTIVES

Gilbert Ahamer, Adrijana Car, Robert Marschallinger, Gudrun Wallentin, Fritz Zobl
Institute for Geographic Information Science at the Austrian Academy of Sciences
ÖAW/GIScience, Schillerstraße 30, A-5020 Salzburg, Austria
gilbert.ahamer@oeaw.ac.at, adrijana.car@oeaw.ac.at, robert.marschallinger@oeaw.ac.at,
gudrun.wallentin@oeaw.ac.at, fritz.zobl@oeaw.ac.at

ABSTRACT
“Perspectives” are seen as the basic element of realities. We propose different
methods to “map” time, sspace, economic levels and other perceptions of
reality. IT allows views on new worlds. These worlds arise by applying new
perspectives to known reality. IT helps to organise the complexity of the
resulting views.

Key Words: Geographic Information Science, mapping, time, space, perception.

0. LET’S START TO THINK 1. WRITING HELPS TO BECOME AWARE

0.1 Our world is the entirety of perceptions. (Our
world is not the entirety of facts.)
Figure 0: The human being perceives the world.
Hence, every individual lives in a different world
(Fig. 0).
0.2 The “indivisible unit”, the atom (ατομος1) of
reality, is equal to one (human) perspective. Our
world is made up of a multitude of perceptions, not
of a multitude of realities and not of a multitude of
atoms (Fig. 1).
We ask: Is it possible to map = write
1. the distribution of material facts and
elements in geometric space? (physics)
2. the distribution of factual events in global
time? (history)
3. the distribution of real-world objects
across the Earth? (geography)
4. the distribution of elements along material
properties? (chemistry)
5. the distribution of growth within sur-
rounding living conditions2? (biology)
6. the distribution of persons acting in
relationships? (sociology)
7. the distribution of individuals between
advantage and disadvantage? (economics)
8. the distribution of perspectives within
feasible mindsets? (psychology)
9. the distribution of living constructs along
selectable senses? (theology)
We see: awareness results from reflection (Fig. 2).
elements
living conditions
objects
personalities

events advantages

matter perspectives
Figure 1: The “primordial soup” of living, before
the advent of (social) organisms: uncoordinated sense
perspectives, uncoordinated world views. x
y
0.3 In order to share one’s own conception with z
space themes time = t
others, “writing” was invented. Similarly, complex
structures, such as landscapes, are “mapped”. To Figure 2: Fundamental dimensions, along which to
map means to write structures. coordinate individual world views when reflecting.
                                                                                                                         
1 2
what cannot be split any further (Greek) životné prostredie (Slovak): living environment

UbiCC Journal – Volume 4 No. 3 609

Special Issue on ICIT 2009 Conference - Applied Computing

2. TIME CAN BE
1. an attribute of space (a very simple
historic GISystem)
2. an independent entity (Einstein’s physics)
3. the source of space (cosmology).
In terms of GIS item 2.1 is expressed as “t is one of
the components of geo-data” i (Fig. 3).

Figure 3: The where-what-when components of

geo-data, also known as triad (Peuquet 2002: 203). Figure 5: Notions of path in a geo-space: (a)
Minard’s map of human losses during Napoleon’s
Time can be understood as 1812 campaign into Russia; and (b) its geo-
• establishing an ordinal scale for events visualisation in a time cube (Kraak, 2009).
• driving changes (= Δ) of realities
• something that unfortunately does not Further examples such as landslides in geology,
appear on paper. growth of plants, energy economics, economics
will be shown in chapter 7.
A proposed solution is to map changing realities
(Δ) instead of mapping time. For implementing the idea to project the t axis onto
Time is replaced by what it produces. This is the Δ axis we need to have clear insight how time
indicated in Fig. 4. quantitatively changes reality.
In other words: we need a model, which (explaining
Δ elements
Δ living conditions how processes occur) determines the representation
Δ objects of time (Fig. 6). Examples are sliding geology,
Δ personalities
ΔGDP/cap, plant growth.
Δ events Δ advantages
One cannot perceive time (never!), only its effects:
Δ matter (e.g. its path) Δ perspectiv. what was perceived in this time span (duration)4?
This is why the t axis is projected onto another axis
Δ sense denoting the effect of elapsed time; what this means
x pro- to the individual sciences is shown in Fig. 4.
y ject!
z
space themes time = t Very similarly, in physics nobody can feel force,
Figure 4: The projection of time (t) onto the effects only its effect (deformation, acceleration), and still
of time (the changes Δ) can apply to any science. forces have been undisputedly a key concept for
centuries.
This idea flips = projects the t axis onto one of the
vertical axes. Time means then: how maps are What is time? Just a substrate for procedures.
changed by the envisaged procedures. What is space? Just hooks into perceived reality.
Such procedures modify the variables along the
axes, be they of physical (gravity force) or of social We retain from this chapter 2 that we need a clear
nature (war). model of how elapsing time changes reality. Then
we can map time as suggested: by its effects.
A classical example is Minard’s map of Napoleon’s
1812 campaign into Russia3 (Fig. 5a, b).
                                                                                                                         
3 4
Patriotic War (in Russian): Отечественная война T. de Chardin’s (1950) concept of durée (French).

UbiCC Journal – Volume 4 No. 3 610

Special Issue on ICIT 2009 Conference - Applied Computing
3. HOW TO WRITE TIME?
The big picture shows us various examples:
1. as a wheel (see the Indian flag): revolving
zodiacs, rounds in stadiums, economic
cycles, Kondratieff’s waves
2. as an arrow (see Cartesian coordinates):
directed processes, causal determinism,
d/dt, d²/dt²
3. as the engine for further improvement
(evolutionary economics): decrease vs.
increase in global income gaps, autopoietic
systems, self-organisation
4. as the generator of new structures
(institution building, political integration,
progressive didactics): new global
collaborative institutions, peer-review,
culture of understanding, self-responsible
learning, interculturality
5. as evolving construct (music).
From this chapter 3 we only keep in mind that the
concepts to understand and represent time are
fundamentally and culturally different.
5
Figure 6: All data representations require models.
4. HOW TO WRITE SPACE?
The big picture shows us various examples:
1. as a container of any fact and any process
(geography and GIS)
2. as result of human action (landscape
planning)
3. as evolving construct (architecture).
Examples span space as
• received and prefabricated versus
• final product of one’s actions, namely:
1. spaces as the key notion for one’s own
science: everything that can be geo-
referenced means GIS
2. space as the product of human activity
3. expanding space into state space: the
entirety of possible situations is
represented by the space of all “state
vectors” which is suitable only if
procedures are smooth.
The main thesis here is: the “effects of time” are
structurally similar in many scientific disciplines,
and they often imply “changes in structures” too.
Information Technology (IT) is already providing
scientific tools to visualise such structures.
5 6
datum (Latin): what is given (unquestionable)
UbiCC Journal – Volume 4 No. 3
5. HOW TO MAP SPACE AND TIME?
The detailed picture: it is obvious that a choice
must be made for one mode of representation and
for one view of one scientific discipline:
1. (x, y; t): cartography, GIS (Fig. 7)
2. (x, y, z; t): geology
3. (x, y, z; vx, vy, vz; t): landslides
4. (x, y, z; biospheric attributes; t): ecology,
tree-line modelling
5. (countries; economic attributes; GDP/cap)
or (social attributes; structural shifts;
elapsing evolutionary time): economic and
social facts in the “Global Change Data
Base”6 (Fig. 8)
6. perceiving rhythms and structures: (only)
these are “worth recognising”: music,
architecture, fine arts.
objects seen by geographers
harmonised
x world views!
y
z
space themes time = t
Figure 7: Harmonising world views: GIS reunites
world views by relating everything to its location.
Different sciences may have considerably different
outlooks on reality (Fig. 8). A humble attitude of
recognising facts5 instead of believing in the
theories one’s own discipline offers can empower
people to survive even in the midst of other
scientific specialties: Galileo’s (1632) spirit: give
priority to observation, not to theories!
This is the essential advantage of geography as a
science: geographers describe realities, just as they
appear. Such a model-free concept of science has
promoted the usefulness of GIS tools to people
independent of personal convictions, scientific
models or theories.
objects seen by economists
harmonised
world views!
x
y
z
space themes time = t
Figure 8: Different but again internally harmonised
world views: explain facts from another angle.
This GCDB is described in Ahamer (2001)
611
Special Issue on ICIT 2009 Conference - Applied Computing
6. WHAT IT DOES, DID, AND COULD DO
6.1 IT helps to organise the multitude of views (=
perceptions) onto data that are generated by
humans:
• IT constructs world views, such as: GIS,
history, economics, geology, ecology etc.
• IT has already largely contributed to
demolishing traditional limitations of space
and time:
o Space: tele(-phone, -fax, -vision), virtual
globes (Longley et al., 2001)
o Time: e-learning, asynchronous web-based
communication, online film storage
(Andrienko & Andrienko 2006).
6.2 This paper investigates non-classical modes of
geo-representation.
We would like to point out that there are two
already well-established fields that offer solutions
to mapping (space and time, Fig. 9) views:
Scientific and information visualisation are
branches of computer graphics and user interface
design which focus on presenting data to users, by
means of interactive or animated digital images.
The goal of this field7 is usually to improve the
understanding of the data presented. If the data
presented refers to human and physical
environments, at geographic scales of
measurement, then we talk about Geovisualisation,
e.g. (MacEachren, Gahegan et al. 2004; Dykes,
MacEachren et al. 2005, Dodge et al., 2008).
Figure 9: Time series and 3 spatio-temporal data
types
(http://www.crwr.utexas.edu/gis/gishydro05/).
6.3 IT could develop tools that are then
interchangeable across scientific disciplines, e.g.
landslides that may structurally resemble
institutional and economic shifts (see 7.1).
IT could prompt scientists to also look at data
structures from other disciplines.
Whatever the disciplines may be, the issues are
structures and structural change!
                                                            
7
http://en.wikipedia.org/wiki/Scientific_Visualization
UbiCC Journal – Volume 4 No. 3
7. EXAMPLES
The authors are members of the “Time and Space”
project at their institution named “Geographic
Information Science”8, a part of which explores the
cognitive, social, and operational aspects of space
& time in GIScience.
This includes models of both social and physical
space and consequences thereof for e.g. spatial
analysis and spatial data infrastructures. We
investigate how space and time are considered in
these application areas, and how well the existing
models of space and time meet their specified needs
(see e.g. Fig. 9). This investigation is expected to
identify gaps. Analysis of these gaps will result in
improved or new spatio-temporal concepts
particularly in support of the above mentioned
application areas.
7.1 Sliding realities: geology
The notion of the path in geography (x, y, t) is
extended by the z axis (see item 5.2) which
produces a map of “time”: Fig. 9 (Zobl, 2009).
Figure 9: Geology takes the (x, y, z; t) world view.
The “effect of time” is sliding (luckily in the same
spatial dimensions x, y, z): we take the red axis in
Fig. 10. Space itself is sufficiently characteristic for
denoting the effects of time.
Figure 10: These effects of time occur in space,
most helpfully. Source: Brunner et al. (2003).
                                                            
8
The overarching aim of the GIScience Research
Unit is to integrate the “G” into Information
Sciences (GIScience, 2009)
612
Special Issue on ICIT 2009 Conference - Applied Computing
7.2 Slices of realities: geology
Despite the lucky coincidence that the effect of
time (Δx, Δy, Δz) occurs in the same space (x, y, z)
we try to produce slides carrying more information
(item 5.3) and hence recur to the so-called attributes
mentioned in Fig. 9 such as grey shades or colours.
The speed of sliding (d/dt x, d/dt y, d/dt z) is
denoted both by horizontal offsets and whitish
colours in the spaghettis (Marschallinger, 2009) of
Fig. 11.
 
Figure 11: The (x, y, z; vx, vy, vz; t) view of a
landslide process (shades of grey mean speed v).
7.3 Slide shows
How to map spatial realities that are not any longer
isotropic displacement vectors of space itself? For
the example of changing tree lines in the Alps
(Wallentin, 2009) a slide show is used to present
the change of growth patterns made up of the
multitude of individual agents (= trees = dots in
Fig. 12). Moving spatial structures are depicted as a
film of structures (item 5.4).
Figure 12: The (x, y, z; biospheric attributes; t)
view of the Alpine tree line (above) and its shift
induced by climate change as a slide show (below).
In such processes which involve independent
behaviour of autonomous agents (here: trees) it
becomes seemingly difficult to apply a
transformation of space itself, e.g. d/dt(x, y, z).
UbiCC Journal – Volume 4 No. 3
7.5 Global deforestation
One key driver for global change is deforestation;
easy to map as change of land use category of a
given area (Fig. 13).
Figure 13: The (x, y, z; Δ biospheric attributes; t):
view of the global deforestation process in mega-
tons carbon. Above: map of carbon flow, below:
time series of GCDB data per nation symbolically
geo-referenced by the location of their capitals.
This representation is analogous to Fig. 11. In both,
the focus shifts from maps(t) Î maps(t, Δt).
Interest includes temporal dynamics:
t = colour (above); Δt = height+colour (below),
enriching the purely spatial interest.
Even if to the aim is to enlarge the scope of the
information delivered from the static map (Fig. 13
above) to the “dynamic map” (Fig. 13 below),
readers will remain unsatisfied because no insight
into the dynamic properties of deforestation is
provided (Fig. 18).
Increasingly, the viewer’s focus turns further from
“facts” to “changes of facts”, to “relationships with
driving parameters9” and to (complex social and
political) “patterns10”.
7.6 Realities beyond slides
But what if the information belongs to the social or
economic realm (Fig. 14)? How to depict economic
levels, education or policies?
Figure 14:
Example for graphic
notation: one
(hypothesised) para-
meter per nation
(seen across the
Jordan = ). 
                                                            
9
see the suggested scenarios for water demand,
water supply and water quality (Ahamer, 2008)
10
Patterns: name of the journal of the American
Society for Cybernetics ASC
613
Special Issue on ICIT 2009 Conference - Applied Computing
7.7 Mapping social processes
Social processes in social organisms can be
described by the intensity of four different
communicational dimensions (Fig. 15) along time:
S = info, A = team, T = debate, B = integration.
This type of writing (Fig. 16) resembles a score in
musical notation11 and was invented for the web-
based negotiation game “Surfing Global Change”
(SGC), its rules are published in (Ahamer, 2004).
The elementary particle of humanity’s progress –
consensus building – is trained by SGC
In this case, IT contributed to making
communication independent from space and time: a
web-platform enables asynchronous worldwide
interaction of participants.
Figure 15: Four basic components of any social
procedure: learning information (Soprano S),
forming a team (Alto A), debating (Tenor T), and
integrating opposing views (Bass B).
Figure 16: A map of social processes in 4 dimen-
sions during a negotiation procedure in a university
course: participants show varying activity levels.
11
partitura (Italian): score (in music)
UbiCC Journal – Volume 4 No. 3
8. TRANSFORMATION OF COORDINATES
8.1 All the above examples have shown that
• various “spaces” can be thought of
• it would be suitable to enlarge the notion
of “time”.
8.2 Suitably, a transformation of coordinates from
time to “functional time” may be thought of.
8.3 In chapter 2, we suggested already to regard
time as the substrate for procedures. Consequently,
different “times” can be applied to different
procedures. As an example, in theoretical physics,
the notion of “Eigentime12” is common and means
the system’s own time.
8.4 Similar to the fall line in the example of
landslides in chapter 7.1 (red in Fig. 10) the
direction of the functional time is the highest
gradient of the envisaged process. This (any!) time
axis is just a mental, cultural construction.
8.5 According to chapter 2 (Fig. 6) a clear under-
standing (mental model) is necessary to identify the
main “effect of time”. We see that such an
understanding can be culturally most diverse. Just
consider the example of economic change:
• optimists think that the global income gap
decreases with development
• pessimists believe that it increases,
hampering global equity.
8.6 Therefore, any transformation of coordinates
bears in itself the imponderability of complex social
assumptions about future global development and
includes a hypothesis on the global future.
8.7 Still, a very suitable transformation is
t Î GDP/capita
(Fig. 17) both because of good data availability and
increased visibility of paths of development.
GDP/cap resembles evolutionary time.
time t GDP/cap
= real time: ≈ evolutionary time of
development:
complex graphic structure simpler graphical structure
Figure 17: A suitable transformation of time uses
the economic level, measured as GDP per capita.
12
literally (German): the own time (of the system)
614
Special Issue on ICIT 2009 Conference - Applied Computing
8.8 The strategic interest of such a transformation is
“pattern recognition”, namely to perceive more
easily structures in data of development processes.
Examples for such “paths of development” are
shown in Fig. 18 for the example of fuel shares in
energy economics.
Figure 18: Structural shift of percentages of
various fuels in all nations’ energy demand 1961-
91. Data source: GCDB (Ahamer, 2001).
8.9 It is suggested here that implicitly during many
mapping endeavours such transformation occurs.
This is legitimate, but care must be taken to take
into account the (silently) underlying model of
human development.
8.10 Suitable transformation of coordinates can
facilitate to see and communicate evolutionary
structures, as it enables common views of humans
and is therefore helpful for global consensus
building.
8.11 Also the “effects of time” are projected into a
common system of understanding which might give
hope to facilitate common thinking independently
of pre-conceived ideologies.
This plan creates the “common reference system of
objects”.
8.12 This paper suggests enlarging the concept of
• “globally universal geo-referencing” (one
of the legacies of IT)
to
• “globally universal view-referencing”
• or “globally universal referencing of
perspectives” 13.
Fig. 19 illustrates this step symbolically.
13
The facts themselves may well be delivered by
endeavours such as Wikipedia but here it refers to
the perspective on facts! A huge voluntarily
generated database on people’s perceptions, views
and opinions would be needed.
UbiCC Journal – Volume 4 No. 3
9. A FUTURISTIC VISION
9.1 Building on the vision of “Digital Earth” (Gore,
1998), the deliberations in this paper might
eventually lead to the vision of “Digital
Awareness”: the common perspective on realities
valid for the global population, aided by
(geo)graphic means.
9.2 The primordial element of (human and societal)
evolution is consensus building. Without ongoing
creation of consensus global “evolutionary time” is
likely to fall back.
The futuristic vision is to map global awareness.
Figure 19: The global society perceives the world.
9.3 Much like the georeferenced satellites which
circulate around the world produce a “Google,
Virtual [or similar] Earth”, the individual spectators
in Fig. 19 circle around the facts – and they create a
“common virtual perception”: an
IIS = Interperspective Information System.
the entirety seen by all
global citizens
x
entirety of
world views!
y
z
time = t
space of themes
Figure 20: Divergent perceptions circulate around
earthen realities. The entirety of world views cre-
ates the IIS (Interperspective Information System).
615
Special Issue on ICIT 2009 Conference - Applied Computing

9.4 Do we just mean interdisciplinarity? No. Nor do 10. CONCLUSION

we simply refer to people looking into any Sciences are similar to “languages” spoken by
direction. Fig. 21 shows the difference to IIS. people, they differ globally. Understanding for
others’ languages is essential for global sustainable
peace.
Human perceptions are also strongly influenced by
underlying models, assumptions and preconceived
understandings.
Studying geo-referenced data sets (GIS) can help to
facilitate bridging interperceptional gaps.

For the transformation of world views – to make

them understandable – it is necessary to know
about
• the “effect of time”, namely the “path along the
continuum of time” which a variable is expected
to take
• the speakers’ underlying model of a complex
techno-socio-economic nature
• the resulting perception of other humans.

A future task and purpose of IT could be to

Figure 21: This is not IIS.
9.5 The science of the third millennium will allow
dealing with a multitude of world views and world
perspectives (see Tab. 1) with an emphasis on
consensus building.
When learning, the emphasis lies on social learning
and may also make use of game-based learning
(such as the web-based negotiation game “Surfing
Global Change”) which allows to experimentally
experiment with world views without any risk
involved.
Table 1: The science of the third millennium
encompasses multiple perspectives
element interaction perspective
single Mechanics Logics
combine the multitude of (e.g. geo-referenced) data
and to rearrange it in an easily understandable
manner for the viewpoints and perspectives of
another scientific discipline or just another human
being. Such a system is called Interperspective
Information System IIS.
Merging a multitude of perspectives to form a
common view of the entire global population is the
target of an IIS.
Symbolically, a “Google Earth”-like tool would
eventually develop into a “Google World
Perspective”-like tool, or a “Virtual Earth”-like tool
would become a “Virtual Perspective” tool
encompassing all (scientific, social, personal,
political, etc.) views in an easily and graphically
understandable manner.
In the above futuristic vision, IT can/should(!)
become a tool to facilitate consensus finding. It can
Teaching
21st cent.

rearrange the same data for a new view.

19th cent.

20th cent.

ones
mani- Thermo- Systems Social Symbolically speaking: similar to Google Earth
fold dynamics analysis learning which allows one to view the same landscape from
gaming, IIS different angles, a future tool would help to
navigate the world concepts, the world views and
9.6 A suitable peaceful “common effort14” for a the world perspectives of the global population.
peaceful future of humankind would involve IT can reorganise extremely large data volumes (if
developing tools and visual aids in order to technological growth rates continue) and could
understand the opinions of other citizens of the eventually share these according to the viewpoint of
globe. the viewer.
The future is dialogue.
Such a step of generalisation would lead from
Or else there will be no future. “Geographic Information Science” to
                                                             “Interperspective Information Science”, implying
14
(jihad in Arabic) also means: common effort the change of angles of perception according to
of a society one’s own discipline.

UbiCC Journal – Volume 4 No. 3 616

Special Issue on ICIT 2009 Conference - Applied Computing

REFERENCES
Ahamer, G. (2001), A Structured Basket of Models
for Global Change. In: Environmental
Information Systems in Industry and Public
Administration (EnvIS). ed. by C. Rautenstrauch
and S. Patig, Idea Group Publishing, Hershey,
101-136, http://www. oeaw-
giscience.org/ProjectFactSheets/Project
FactSheet_GlobalChange.pdf.
Ahamer, G., Wahliss, W. (2008), Baseline
Scenarios for the Water Framework Directive.
Ljubljana, WFD Twinning Project in Slovenia,
http://www.oeaw-giscience.org/ProjectFact
Sheets/ProjectFactSheet_EU_SDI.pdf.
Andrienko, N., Andrienko G. (2006), Exploratory
Spatial Analysis, Springer
Brunner, F.K., Zobl, F., Gassner, G. (2003), On the
Capability of GPS for Landslide Monitoring.
Felsbau 2/2003, 51-54.
de Chardin, T. (1950), La condition humaine [Der
Mensch im Kosmos]. Beck, Stuttgart.
Dodge, M., McDerby, M., Turner, M. (eds.) (2008)
Geographic Visualisation, Wiley
Dykes, J., A. MacEachren, et al. (2005). Exploring
Geovisualization. Oxford, Elsevier.
Galileo, G. (1632), Dialogo sopra i due massimi
sistemi del mondo, tolemaico, e copernicano.
Fiorenza.
GIScience, (2008), Connecting Real and Virtual
Worlds. Poster at AGIT’08, http://www.oeaw-
giscience.org/index.php?option=com_content&ta
sk=blogcategory&id=43&Itemid=29.
Gore, A. (1998). Vision of Digital Earth,
http://www.isde5.org/al_gore_speech.htm.
Kraak (2009), Minard’s map.
www.itc.nl/personal/kraak/1812/3dnap.swf
Longley, P.A. et al. (2001) Geographic
Information. Science and Systems, Wiley
MacEachren, A. M., M. Gahegan, et al. (2004).
Geovisualization for Knowledge Construction
and Decision Support. IEEE Computer Graphics
& Applications 2004 (1/2): 13-17.
Marschallinger, R. (2009), Analysis and Integration
of Geo-Data. http://www.oeaw-giscience.org/.
Peuquet, D. J. (2002). Representations of Space and
Time. New York, The Guilford Press.
Wallentin, G. (2009), Ecology & GIS.
Spatiotemporal modelling of reforestation
processes. See http://www.oeaw-
giscience.org/images/stories/Downloads/pecha%
20kucha%20technoz%20day.pdf
Zobl, F. (2009), Mapping, Modelling and
Visualisation of georelevant processes.
http://www.oeaw-giscience.org/.
                                                            
i
GIScience goes way beyond this view of time and
space (considering time as function) because it al-
lows for much more complex queries and analyses.

UbiCC Journal – Volume 4 No. 3 617

 UbiCC Journal
Conference Consulting

UbiCC Journal Conference

Consulting

UbiCC Journal exhibits many

conferences a year. The calendar of
conferences includes Trade Shows,
Book Fairs, Conferences and Exhibits
throughout the world.

From local meetings to international

conferences, UbiCC offers you access
to a conference community that will
help you meet your professional goals.
UbiCC sponsors multiple conferences
and meetings and is involved in the
technical program development of
events around the world.
.

simplifying IT consulting
Conference
Management
support

UbiCC Journal
Conference Consulting
4 Nichols Rd, Nesconset, NY
+1-347-4149239 ph
+1-212-901-6990 fax
www.ubicc.org
 Technology at work for you
CONNECTING YOUR CONFERENCE TO THE TECHNOLOGY RESOURCES YOU NEED

Network with your peers and enhance your career. links to conference websites, and special issue with A B O U T US
Whether looking for your industry's hottest topics, the UbiCC for technical program, registration and The papers of original research

latest research breakthroughs and innovations, short
courses on changing technologies, or vendor exhibits
of new commercial products and services, UbiCC has
it covered.
Find the conferences that meet your needs with the
UbiCC Conference Search, where you can search by
additional information. and innovatory applications from all
parts of the world are welcome.
Conferences offer a unique opportunity for developing
The papers for publication in
a wide range of professional relationships. Expert UBICC are selected through
panels, presentations, key note speeches, exhibits rigorous peer reviews to ensure
originality, appropriateness,
and social events all provide an opportunity to interact
significance, and readability.
with the leaders and innovators in your field of interest.

date, location, or UbiCC sponsor. Search results

I T S U P P O RT S O L U T I ON S

• CONFERENCE MANAGEMENT SYSTEM

flexible solutions for C O N F E R E NC E C O NS UL T I N G P R OV I DE S

• JOURNAL MANAGEMENT SYSTEM
• PAPER REGISTRATION SYSTEM
• CONFLICT RESOLUTOR
your conference needs A TOTAL END TO END SOLUTION.
SCOPE
The Journal aims to provide an
environment in which ideas and
P R E SE N T AN D P UB L ISH A P P LI C A TIO N M A NA GE M E N T research can be presented,

• ANTI PRAGULISM SOLUTION discussed and criticized so that best

• UBICC XPLORE SEARCH
• PAPER TRACKER
• UBICC PERSONALIZED TOOLBAR
Whether your paper is focused on research and UbiCC Journal conferece collaboration solution is a free practice can be assimilated into the
conference management system that is flexible, easy to new curricula of schools, colleges
development, or engineering and implementation, UbiCC
use, and has many features to make it suitable for and universities. For further details
has the right forum for your work. For upcoming various conference models. It is currently probably the contact: info@ubicc.org
submission opportunities, check the Call for Paper most commonly used

deadlines. conference management

• Selected papers from UbiCC authorized conferences system.

Journal/Conference
handling solution has been
BE A LEADER designed to help program
chairs to cope with the
complexity of the
As an UbiCC conference organizer you can connect with
refereeing process. The E D I T O R IA L B O A R D
others in your profession, make new contacts, and
current version supports: The editorial board of Ubiquitous
contribute to your field of interest. computing and communication
Have you identified a need for a conference in an • Sophisticated and flexible management of the access Journal consists of leading specialists
of PC members and referees; of the field from the globe.
emerging technology, looking to help bridge the gap
• Online discussion of papers;
between research and development and engineering
• Monitoring emails;
and implementation? Become a conference organizer • Paper submission;

and make a difference. and many more …