Sunteți pe pagina 1din 49

Environment, Computing and Energy Efficiency:

How Virtualization Can Contribute to Reduce the


Carbon Footprint of Today's Computing

Mohsen Mojabi

Submitted to the
Graduate Committee of School of Computing and Technology
in partial fulfilment of the requirements for the Degree of

Master of Technology
in
Information Technology

Eastern Mediterranean University


February 2014
Gazimausa, North Cyprus

Approval of the Graduate Committee of School of Computing and Technology

Assoc. Prof. Dr. Mustafa lkan


Director

We certify that we have read this project and that in our opinion it is fully adequate in scope and
quality as a project for the degree of Master of Technology in Information Technology.

Asst. Prof. Dr. Ersun iolu


Supervisor

Examining Committee
1. Assoc. Prof. Dr. Mustafa lkan
2. Asst. Prof. Dr. Alper Doanalp
3. Asst. Prof. Dr. Ersun iolu

ABSTRACT

The mankind has faced many new phenomena over the course of last couple of
decades. One, for sure, is the IT revolution. One cannot even imagine a world
without computers on each desk nowadays; and thats not the whole picture. The
need for huge back-end computing has led to proliferation of huge data-centers.
Although the move towards Cloud Computing has already been started for couple of
years now, there are still so many companies and large enterprises which rely solely
on their own server rooms. But what about the other phenomenon the world is just
facing nowadays: Global Warming as a consequence to the increase in carbon
dioxide emissions created by power plants. Everyday scientists and manufacturers
are working hard to find a new method/technique to reduce the energy consumption
of consumer electronic devices, in order to reduce what is known as carbon
footprint of that device, but what about desktop computers and servers? Aside from
all thats been done for decreasing their energy consumption, is there a new way to
progress even further down this road? In this project, we are going to discuss how
virtualization can influence the new way of computing, and how drastically it can
decrease the carbon footprint of computing, both on front-end (Desktop
Virtualization) and back-end (Server Virtualization).

Keywords: Virtualization, Server Virtualization, Hardware Virtualization, Green


Computing, Energy Consumtion Efficiency, Carbon Footprint, Carbon Footprint
Reduction, Global Warming, Computing Environmental Impact.

ACKNOWLEDGMENTS

I would like to thank Asst. Prof. Dr. Ersun Iscioglu for his continuous support and
guidance in the preparation of this study. Without his invaluable supervision, all my
efforts could have been short-sighted.

I also would like to thank Assoc. Prof. Dr. Mustafa lkan, Director of the School of
Computing and Technology, Eastern Mediterranean University.

II

To my late father

III

TABLE OF CONTENTS

ABSTRACT ............................................................................................................. I
ACKNOWLEDGMENTS........................................................................................ II
DEDICATION .......................................................................................................III
TABLE OF CONTENTS........................................................................................IV
LIST OF TABLES................................................................................................... V
LIST OF FIGURES ................................................................................................VI
LIST OF ABBREVIATIONS ............................................................................... VII
1 INTRODUCTION .................................................................................................1
2 WHAT IS CARBON FOOTPRINT AND HOW IT SHOULD BE MEASURED...5
3 TRADITIONAL APPROACHES VS. VIRTUALIZATION ................................13
3.1 VIRTUALIZATION .....................................................................................15
3.1.1 SERVER (HARDWARE) VIRTUALIZATION .....................................16
3.1.2 NETWORK FUNCTION VIRTUALIZATION......................................17
3.1.3 STORAGE VIRTUALIZATION............................................................17
3.1.4 DESKTOP VIRTUALIZATION (VDI)..................................................18
3.2 ADVANTAGES OF VIRTUALIZATION ....................................................19
3.3 DESKTOP VIRTUALIZATION AND ITS FUTURE...................................20
4 CASE STUDY.....................................................................................................21
5 CONCLUSION ...................................................................................................33
REFERENCES

..................................................................................................35

IV

LIST OF TABLES

Table 4.1: Definition of different business sizes based on number of employees .....22
Table 4.2: Pre-virtualization total carbon footprint ..................................................29
Table 4.3: Server consolidation ratios for each business type ..................................29
Table 4.4: Post-virtualization total carbon footprint.................................................30

LIST OF FIGURES

Figure 2.1(a): U.S. Total GHG Emissions .................................................................8


Figure 2.1(b): Emissions by Economic Sector...........................................................8
Figure 2.2: Total carbon footprint of the Dell PowerEdge R710 ..............................10
Figure 2.3: Personal computing related electricity vs. global electricity use.............11
Figure 2.4: Sector-wise electricity consumption ......................................................11
Figure 2.5: Global electricity consumption vs. total ICT sector consumption ..........12
Figure 4.1: Power consumption results of an HP ProLiant ML110 G5 ....................24
Figure 4.2: Power consumption results of an HP ProLiant DL380 G7 .....................25
Figure 4.3: Power consumption results of an HP ProLiant DL580 G5 .....................26
Figure 4.4: Power consumption results of an HP ProLiant BL280 G6 .....................27
Figure 4.5: Results for Small/Medium/Large Businesses.........................................31
Figure 4.6: Results for Enterprise Business .............................................................31
Figure 4.7: Total Carbon Footprint Reduction in Each Business Type.....................32

VI

LIST OF ABBREVIATIONS

3D

Three Dimensional

ACPI

Advanced Configuration and Power Interface

APM

Advanced Power Management

ARM

ARM Holdings (Company Name)

ATA

Advanced Technology Attachment

CISC

Complex Instruction Set Computing

CO2

Carbon Dioxide

CO2eq

Carbon Dioxide Equivalent

CPU

Central Processing Unit

DPMS

Display Power Management Signaling

EPA

Environmental Protection Agency

GHG

Green House Gas

GHz

Giga Hertz

HDD

Hard Disk Drive

HP

Hewlett-Packard (Company Name)

I/O

Input/Output

IBM

International Business Machines (Company Name)

ISO

International Standardization Organization

IT

Information Technology

KB

Kilo Bytes

kWh

Kilo watts per hour

MB

Mega Bytes

MHz

Mega Hertz

MWh

Mega Watts per hour

VII

NFV

Network Function Virtualization

OS

Operating System

PC

Personal Computer

RAM

Random Access Memory

RISC

Reduced Instruction Set Computing

ROI

Return on Investment

SAN

Storage Area Network

SDN

Software-Defined Networking

SMB

Small-to-Medium Business

SPEC

Standard Performance Evaluation Corporation

TIA

Telecommunications Industry Association

UK

United Kingdom

US

United States

VDI

Virtual Desktop Infrastructure

VESA

Video Electronics Standard Association

vs.

versus

VIII

Chapter 1

INTRODUCTION

Computers have been around for decades, but the real boom in computing began
with the introduction of personal computers, in which brought computers to everyday
life of people. With the advancement of technology, computers become cheaper and
widely available, and at the same time evolved from sparsely isolated standalone
personal computers to an internetwork of connected nodes. Upon the invention of the
Internet and its related technologies like Web, the need for computers grew, and now
theyve become an inseparable part of our lives. Gartners statistics show that the
billionth personal computer has been delivered in April 2002 [1]. The second billion
personal computer has probably shipped back in 2007. A report by Forrester
Research claims that by the end of 2008 there were at least a billion personal
computers in use (worldwide), and it is growing fast especially in the emerging
markets. Estimates by Forrester show that at least two billion personal computers
will be in use by 2015 (not taking into account mobile devices and tablets) in use by
2015. Hence, although it took 27 years for the world to reach one billion personal
computers in use, it would take only 7 years for to reach to 2 billion. But this is not
the whole image. Approximately 8 to 8.5 million rack and blade servers are shipped
annually [2]. Facebook, Google, Rackspace and other web giants are now building
their own servers so the numbers will be skewed upwards lightly [3]. With an
average life expectancy of 5 years, one can estimate there are about 45-50 million
servers in operation in the world now. Businesses now heavily rely on computing

systems, and more computational power is needed as businesses and manufacturers


grow and leverage new tools and software which aid design and manufacturing.
Now, lets draw another picture. The planet earth is getting warmer and it is
mostly blamed on Green House Gas (GHG) emissions and it becomes more severe
every day. Its impacts include extreme weather conditions, earthquakes, droughts,
glacier retreat, tsunamis, rising sea levels, increased volcanism, ocean acidification
and temperature rise and oxygen depletion [4].
World Bank 2012 report[5] states planet will be 4 degrees warmer by 2100.
According to Jim Yong Kim (World Bank Group president) comparing to preindustrial era, the earth is now 0.8 warmer [6]. Given the above figures, its not a
surprise that all industries are trying to reduce the energy consumption of their
products, and computer industry is no exception. A term which is widely used
nowadays is Carbon Footprint, the carbon dioxide quantity produced by a particular
entity as a result of fossil fuel consumption [7]. The computer industry has taken
measures to reduce the carbon footprint of computers and related devices by making
them more efficient and leveraging energy saving techniques. One of the earliest and
most significant attempts in early 1990s can be considered a program called the
ENERGY STAR [8]. Although these efforts have contributed to the reduction of the
energy consumption, but the main obstacle is in fact underutilization of the available
computing power. Computers are ordered based on the needs, but during their life
time, there are rare moments in which theyre computational power is fully utilized,
and most of the times the workload on the CPU is a fraction of their capability. This
phenomenon is seen both on Clients (Business Desktop Computers) and Servers.
Mechanisms such as Dynamic CPU Frequency [9] are developed to address this issue
and reduce the energy consumption in periods which the CPU is under relatively low

utilization, but the most effective technology so far seems to be [Hardware]


virtualization. It pertains to the creating virtual machines which resembles a physical
computer along with its OS [10]. Applications run on a VM are segregated from the
hardware underneath [10]. Using this technology, the physical box is being used as a
host to the many virtual machines and to enable this, a special type of software is
being used which is commonly referred to as Hypervisor [10]. In order to
differentiate the software which consists the VM (The real operating system) from
the software which provides the simulated version of hardware resources to the VMs
(Hypervisor), usually the terms guest host are used. The hypervisor can be
deployed for Server Virtualization, in which a group of virtual Server Machines are
being hosted (consolidated) on fewer physical servers, and it also can be deployed in
Desktop Virtualization [11] in which virtual PCs are being hosted on much fewer but
stronger physical machines and the client (commonly referred to as Thin Client) is
only used as an interface connecting the user to the virtual machine running at the
server-side just like the old concept of mainframes and terminals.
In this research, were are going to first brief on how carbon footprint is defined
and measured, discuss the different virtualization technologies, and then measure the
effectiveness of virtualization in reduction of the carbon footprint of computing by
creating four different scenarios ranging from small businesses to enterprise scale
business and then incorporating real-world data into it. For this purpose, first sample
business will be depicted and measured by using conventional physical systems, and
then they will be migrated to virtualization (virtualized). In this study the focus is not
only on the CO2 produced as a consequence of electricity consumption by the
computing devices, but also the role that manufacturing and then (after its end of life)
decommissioning and recycling processes plays in production of carbon dioxide

emissions. Therefore, as well as calculating carbon dioxide emissions of energy


consumption

by

the

hardware,

the

carbon

dioxide

emissions

for

the

manufacturing/decommissioning/recycling of the hardware will also be taken into


account.

Chapter 2

WHAT IS CARBON FOOTPRINT AND HOW IT


SHOULD BE MEASURED

A quick research in scientific literature, publications and other media in general,


it is revealed that 'carbon footprint' as a term has turned into a buzzword, however
without being clearly defined. The term is now widely used in the public discussions
on responsible action against global warming and climate change, in which it is
frequently heard in government speeches, in business domain and in the media.
But there is lack of a clear and standard definition of this term and how exactly
carbon footprint should be measured. Several attempts had been made to suggest a
definition that all can agree upon. It seems that the term carbon footprint has its
roots in the term Ecological Footprinting [12]. Commonly it is perceived that it
pertains to the amount of emitted gasses as a consequence of production or
consumption activities which result in climate change. Also there is no consensus on
how carbon footprint should be measured or quantified. Suggestions range from
measuring total carbon dioxide produced to measuring whole life-stage Green House
Gasses. Also there is no general agreement over what unit of measurement should be
used. Some examples are given below:

"A carbon footprint is a measure of the amount of carbon dioxide emitted through
the combustion of fossil fuels. In the case of a business organization, it is the amount
of CO2 emitted either directly or indirectly as a result of its everyday operations. It

also might reflect the fossil energy represented in a product or commodity reaching
market."[13]

The carbon footprint was calculated by "measuring the CO2 equivalent emissions
from its premises, company-owned vehicles, business travel and waste to
landfill."[14]

"The carbon footprint is the amount of carbon dioxide emitted due to your daily
activities from washing a load of laundry to driving a carload of kids to
school."[15]

"Carbon footprint is the total amount of CO2 and other greenhouse gases, emitted
over the full life cycle of a process or product. It is expressed as grams of CO2
equivalent per kilowatt hour of generation (gCO2eq/kWh), which accounts for the
different global warming effects of other greenhouse gases."[16]

"Carbon Footprint is a measure of the impact human activities have on the


environment in terms of the amount of greenhouse gases produced, measured in tons
of carbon dioxide."[17]

The definition which is suggested by Thomas Wiedmann and Jan Minx (from
Stockholm Environment Institute, University of York, UK), is "Carbon footprint is a
measure of the exclusive total amount of carbon dioxide emissions that is directly
and indirectly caused by an activity or is accumulated over the life stages of a
product"[18]. This one seems to be the definition that would serve best to the scope
of this study since it only takes into account the carbon dioxide as the element to be

measured (and not taking into account the other Green House Gasses), and also it
considers not only the activity of a certain entity (the amount of electrical power
used) but its life stages as well. When speaking of life stage, it is important to put
focus on the parts where GHG emissions are inevitable. Take Wind Turbines as an
example. One would think as they generate electricity by wind, they would have no
carbon footprint at all. But that is not true. The amount of GHG a wind turbine
produces in its lifespan (its carbon footprint) is approximately 11 grams
(CO2eq/kWh) [16]. Although the by the normal electricity generation phase it emits
no GHGs, but the manufacturing processes of it do results in GHG emissions,
needless to say the installation and maintenance. Regular inspections and
maintenance implies transportation that is powered by fossil fuels. 98% of the GHG
which is emitted by a wind turbine over a lifespan of 20 to 25 years is made during
the production, assembly and installation of it. 2% is produced by periodic
inspections and maintenance [16]. Wind turbines installed inside the seas and away
from shoreline will create more GHG than onshore ones since it is harder to reach
them [19]. Also the reason that in this study the focus has been put on CO2 emission
only (and neglecting other GHGs) is that firstly, it is the most dominant one (84
percent of all GHG emissions) among the other Green House Gasses (Nitrous Oxide,
Methane and other gasses)[20] (Fig. 2.1a) and it also is the direct consequence of
burning fossil fuels for generating electricity (33 percent of all GHG emissions) and
transportation (28 percent) [21] just as it can be seen in the Fig. 2.1b.

Fig. 2.1a&b: U.S. Total GHG Emissions and Emissions by Sector (2011) [21]

Although there are cases (e.g. DreamHost [22]) in which all the required electricity
by a certain data-center/service-provider is generated by clean sources (e.g.
hydroelectric sources or wind) or so-called sustainable/renewable energy, or
nuclear energy (although some debate it as fully clean energy due to the risk of
radioactive material leakage), for most part electricity is generated by burning fossil
fuels (natural gas, diesel or coal). According to EPA (United States), in 2011 clean
energy accounts approximately for only %13 of the generated electricity (renewable
energy sources 5.6 percent and Nuclear Electricity 7.3 percent) [20]. According to
the Global e-Sustainability Initiative, nearly 2 percent of Global GHG emissions
alone is a direct consequence of the Technology Sector, and of that, Data Center
around the world cause 17 percent of that [23]. Also since in this study the focus is
not only on the power consumption, but all the activities related to manufacturing,
transportation and installation and (when the hardware reached its end of life stage)
decommissioning and recycling, a method needs to be chosen for measuring or at
least estimation- of the carbon footprint related to this part. As it was stated before,
measuring the carbon footprint of a certain computing device needs to take into

account all of the processes involved throughout the whole lifecycle of that specific
computing device (and this can be very hard or next to impossible, at least for the
author), the study will be based on the informations from the manufacturers
themselves. A study by Christopher Weber of Carnegie Mellon University [24] tries
to depict the doubtfulness in evaluation of carbon-footprint by analyzing a server
from IBM confirms that assessing carbon footprint cannot be done precisely. The
study shows that doubtfulness starts from approximately more than fifteen percent
for the manufacturing and transportation stage to more than thirty five percent for the
whole life-stage carbon foot-print [24]. Having said that, not all server manufacturers
conduct such analysis; in fact, as of this writing only two analysis were found in
which the carbon footprint of a piece of hardware is measured, and only for a
certain model. Hence there is no way other than to rely solely on the information
extracted from these two rare cases. Of one of these rare cases, Dells analysis can be
given as an example. In this study [25], the carbon footprint of a Dell PowerEdge
R710 Rack Server has been measured taking into account all aspects, from
development and design to production, customer use and operation to
decommissioning and recycling using ISO 14044/14040 (international standards for
assessment and investigation of the ecological effects a certain product has
throughout its lifespan). Aside from how precise this study can be, its results showed
that 10 percent of the total GHGs produced in the cradle-to-grave GHG emissions are
consequences of the manufacturing, transportation, installation and recycling of the
server (and not its operation during its lifetime in which it consumes electricity over
the course of its lifetime estimated at four years of non-stop operation [25].

Fig. 2.2: Total carbon footprint [kg CO2eq] of the Dell PowerEdge R710Error!
Reference source not found. [25]

Therefore based on the abovementioned research, in this study 10 percent extra


carbon footprint will be considered for the computing equipment (servers) over their
4 years of non-stop operational life time.

How much todays computing contributes to the Global Warming


A study published in 2011 [26] has evaluated the global energy consumption of
computing devices.

10

Fig. 2.3: Personal computing related electricity vs. global electricity use [26]

Fig. 2.4: Sector-wise electricity consumption (million MWh) [26]

As shown in the Fig. 2.3 and 2.4, in 2010 the combined electricity the whole general
computation domain consumed was 1078 million megawatts per hour. Therefore,
computational devices (including other ICT sector devices such as mobile devices)
consumed 6% of the global electricity consumption (Fig. 2.5).

11

Fig. 2.5: Global electricity consumption vs. total ICT sector consumption [26]

And the abovementioned study has only taken into account the electricity
consumption itself. It can be estimated the processes involved in production of
computer hardware devices ranging from material extraction, manufacturing,
transportation to operation, decommissioning and recycling can be attributed to
around 10 percent of the total Green House Gas emissions. This clearly shows the
need for improving the efficiency and reducing the environmental adverse effects. In
this study, it will be shown that by using virtualization and its related technologies
how much can be done towards reaching this goal.

12

Chapter 3

TRADITIONAL APPROACHES VS. VIRTUALIZATION

As mentioned earlier in the first chapter, many attempts have been made to lower
the energy consumption and hence carbon footprint of computing devices. One of the
earliest and most significant is ENERGY STAR program [8] created by Department of
Energy of USA and EPA in the early 1990s, which has also been accepted and used
by EU, Taiwan, Japan, Canada, New Zealand and Australia since then. Computer
products compliant with the ENERGY STAR guidelines commonly consume less
electricity up to thirty percent [27]. Many standards and protocols have been
developed around the idea of ENERGY STAR, e.g. VESA DPMS which defined
mechanism for turning on and off the monitor and thus enabling the PC to take
control over the power management of the monitor and be able to switch off the
display after a certain period of being idle. This approach led to introduction of the
next generation of these sorts of technologies, namely APM (Advanced Power
Management) and later, ACPI (Advanced Configuration and Power Interface). APM
was first introduced in 1992 by Intel and Microsoft [28], and later has been upgraded
to ACPI in 1996[29]. ACPI which was later introduced by a cooperation between
Microsoft, Phoenix, Intel, Toshiba and HP, and gained wider adoption with many
operating systems and processor architectures.

Other mechanisms which were developed to reduce power consumption in


computers include low power memory modules and Dynamic CPU Frequency
techniques. Low power memory modules can operate with up to 35.5 percent lower

13

power [30]. CPU throttling (or Dynamic frequency scaling) is a mechanism in which
the microprocessor frequency is adjusted automatically, to save electricity as well as
to decrease the chips temperature [9]. Less heat generated enables the system fans to
operate with lower speeds or even be switched off, in which it allows more reduction
in electricity consumption. In servers which should be operational continuously and
hence ACPI cannot be applied, when the server is underutilized, Dynamic CPU
frequency helps to decrease the total consumption

of energy, both by CPU

electricity consumption reduction and also by reducing the heat generated (which
needs more cooling).

Another technology which was around for decades but recently has made its way
first to handheld devices and then surprisingly- to server domain, is the RISC
architecture. Historically, CPU architecture is divided into either CISC or RIS. They
both were used in CPU design quite extensively (each has their own characteristics),
until a British company named ARM [31] started to develop processors by RISC
architecture which uses very-low power and therefore generates very low heat
(which eliminates the need for conventional cooling methods, e.g. heat-sinks, fans
and ventilation). These features made them the best option primarily for mobile and
handheld devices. Unlike Intel processors which can consume up to 165 Watts (e.g.
Intel Xeon 7000 Series)[32] and therefore require extensive cooling, ARM
processors usually consume 4 Watts or less and need not much of cooling. As
recently their computational powers increased recently to more than 2 GHz sporting
4 cores [33] and more, some server manufactures have started to create very-low
power-consuming servers by using ARM processors. For instance, HP ProLiant
Moonshot enterprise server series consume 89 percent less energy than a regular

14

server with almost the same capability (A simple Intel Atom-based server the with
same capability consumes usually 40 Watts in contrast to nearly 20 Watts for a
typical HP ProLiant Moonshot server [34]).Although these servers cannot be used
for very CPU intensive workloads such as 3D modelling and database processing,
but for light loads like static webpage hosting can be very beneficial. But still the
issue of underutilization is left somehow unaddressed. None of the abovementioned
technologies can fully address the fact that most servers are not operating at their full
power, and although at peak times their utilization and load may peak up to 90
percent, for most of the times theyre staying either idle or heavily underutilized,
operating at less than 5-7 percent of their actual capability [35] .The underutilization
not only happens at the processing (CPU) part, but also at their memory and hard
disk space. HDDs come in certain sizes and their physical size cannot be shrunk or
extended on demand. So, based on the size of the used HDD and the installed OS
requirements, the HDD is never fully utilized, and the extra space is actually wasted
and will never be used; also this extra space cannot be temporarily allocated to any
other system for temporal uses of other systems when they are in need of it. The
Virtualization technology has been developed to answer to all these
underutilization issues.

3.1. Virtualization
By definition, virtualization is a way of creating a virtual entity in a way in which
it resembles the actual hardware resource. By using virtualization, the actual
hardware resource can be divided into one or many virtual instances of that certain
piece of hardware [36]. Traditionally, Hardware and Software are braided together
and cannot be segregated. For instance, a piece of software is usually being executed

15

on a certain CPU, being allocated a certain amount of RAM, residing on a certain


space of disk and communicate using specific I/O ports. By using virtualization,
simply these relationships are broken and an abstraction layer is created between the
software and the hardware. By using virtualization, software is decoupled from the
hardware and the resulting flexibility makes the possibility to move workloads and
consolidate them together in order to increase the utilization of hardware. By using
Virtualization, multiple workloads can coexist on the same hardware, in which it
enables hardware aggregation and increased utilization. Virtualization enables VMs
to be moved between physical hardware resources without any or minimaldisruption. For instance, a VM can be moved seamlessly to another physical host
enabling the first host to be brought offline either for maintenance or replacement.
VM consolidation reduces hardware needs and therefore the total cost of ownership.
As virtualization can happen at many levels, below a number these technologies are
explained:

3.1.1. Server (Hardware) Virtualization


Server (or Hardware) virtualization refers to segmentation of a (physical) server
into several virtual ones in order to increase hardware utilization [37]. By this
technique, special software (called Hypervisor) divides the different resources of the
physical server (Disk, RAM, CPU, etc.) into multiple virtual ones. To refer to the
VMs running on top of Hypervisor and to the Hypervisor running on top of the
physical servers bare-metal, terms host and guest are used respectively. Server
(Hardware) virtualization proved to have number of advantages; as it allows each
VM run its own OS and also every VM can be rebooted independently of the others.
This technology reduces total cost of ownership due to minimizing the need for

16

hardware. Server virtualization also conserves data center rack space through
consolidation several servers into reduced number of physical boxes. Less physical
boxes equals to less required hardware maintenance and lower costs.

3.1.2. Network Function Virtualization


Network Function virtualization (NFV) is using network resources through a
logical segmentation of a single physical network[38]. Network virtualization refers
to creating an aggregated logical pool of networking resources consisting of the
actual networking equipment, accessible regardless of its physical criteria [39].

3.1.3. Storage Virtualization


Storage virtualization refers to consolidation of several different networked
storage systems into logically an individual storage entity [40].

Storage

virtualization is commonly leveraged in Storage Area Networks. It facilitates and


quickens storage-related workloads (e.g. back-up and archiving) [41].

3.1.4. Desktop Virtualization (VDI)


Desktop virtualization segregates the desktop environment from the physical
hardware required for it [42]. It somehow resembles the good old mainframe days of
centralized computing while providing the similar functionality to traditional user
experience of desktop computing. Every user will have their own isolated copy of a
desktop operating system and related applications, but actually it is a virtual machine
on a host server, which the user is able to access it via a low-cost thin client (some of
them are called zero clients, since they provide extremely limited features, almost
nothing but the connectivity, and theyre relatively cheap) just like a mainframe-era
terminal.

17

This way, all those desktop user environments can be managed from one central
management console. Installing new software/upgrades and security hotfixes will be
done with way less administrative overhead. And the risk of user behaviors and
mistakes in which it can breach security and/or cause unplanned downtimes drops
dramatically.
Similar to server virtualization, desktop virtualization is built on top of a
hypervisor, which serves on hardware and builds a platform on which IT staff can
deploy and centrally administer users virtual desktops. Desktop virtualization
essentially provides each user a virtual machine which contains a separated and
isolated copy of the users operating system and his/her required applications
installed.

3.2. Advantages of virtualization


Virtualization offers capabilities and efficiencies that are not feasible in the
physical- server-only realm, from reduced costs, faster provisioning to saving energy.
Main advantages can be categorized as below:

Server consolidation: Software serving on different servers usually dont


fully utilize the resources available to them, thus by using virtualization more
servers can be run on less servers. Estimations show that physical servers are
commonly running at less than twenty percent utilization. By using
virtualizing, enterprises can utilize their available hardware between 60
percent and 80 percent [43]. By definition, "Consolidation Ratio" is the
number of virtual servers which can be hosted on each physical host. Taking
the above mentioned example, it means a 3 or 4 consolidation ratio,

18

respectively. Based on the hardware capacity of the physical server, a


virtualization host can sometimes reach consolidation ratios up to 20 [44].

Smaller footprint: Virtualizing servers will result in reduction of many


separate physical servers and hence reduces the number of actual physical
boxes. This will also result in a more compact datacenter, reduction in
cooling needs and hence decrease in both electricity costs and carbonfootprint.

Hardware costs: Again as stated above, since virtualization enables


increased utilization of available hardware resources and hence decreases the
need for physical servers (the phenomenon of inevitable physical server
proliferation is also known as Server Sprawl[45]), it will increase the total
cost of ownership both by reducing need for hardware which in turn reduces
the cost of maintenance.

3.3. Desktop Virtualization and its future


VDI offers desktop computing a more agile, flexible and resource-efficient way
than traditional solutions. Virtual desktops may replace the traditional desktop by
providing the same experience and power. The flexibility of VDI enables each user
to have his/her personal computer configured in a certain way tailored to meet his/her
needs, no matter how memory/disk/processor intensive it might be.
VDI changes the desktop computing structure from individual devices carrying
all of the required computing power to another system in which all of the required
computing power is centralized in a data center. Since local devices dont require
much computational and processing power, lower-cost devices called thin clients
can be used which consume so much less energy (7-15 Watts [46] in comparison to

19

80-110 Watts for a normal PC[47]) and therefore create much less carbon footprint.
These devices have significantly prolonged lifespan than PCs, because when there is
a need to upgrade systems and provide more resources (Disk/CPU/RAM) to each
user, there will be no need to replace the thin clients; instead the resources for each
user will be increased at server side. This approach again eliminates the need for
frequent hardware refreshment and therefore the carbon-footprint related to it [48].

20

Chapter 4

CASE STUDY

The objective for this study was to emphasize considering the Carbon Footprint
of computing in general as opposed to only considering the monetary value of the
consumed energy by computing hardware, and showing that virtualization can
contribute to this aim. As of this writing, no similar researches have been found to
look at virtualization from a Carbon Footprint perspective, and measure the
effectiveness of Virtualization in reduction of the Carbon Footprint. In previous
chapters, first a certain definition for carbon footprint was elected, and then
currently-available information concerning measurement of carbon footprint for the
whole life stage of a certain computing device was provided. Due to utter scarcity
of information regarding whats known as cradle-to-grave carbon footprint of
computing devices, as for this Term Project, the predictions will be based solely on
the Dells study. As mentioned earlier in Chapter 2, a study [25] done by Dell in
which all aspects from development and design to production, customer use and
operation to decommissioning and recycling using ISO 14044/14040 is taken into
account, showed that in 4 years of being constantly operational, around 10 percent of
the total carbon footprint produced by a typical server in US is exclusively produced
by

the

manufacturing,

transportation,

installation

and

then

the

decommissioning/recycling phase and not by the operation of the server itself. This
information will be the base of the calculations in this chapter for taking into account
the non-operational carbon footprint of a typical servers cradle-to-grave total
carbon footprint.

21

Methodology:
In this chapter, for measuring the effectiveness of virtualization in reducing the
carbon footprint of the computing, firstly the computing needs of the 4 typical
companies from different sizes ranging from small to enterprise- will be assumed,
their IT infrastructure schema will be first depicted based on traditional nonvirtualized systems, the carbon footprint will be measured for 4 years of operation
(the extra 10% non-operational phase carbon-footprint will be calculated and kept
for further calculations as well), and then all of the servers will be virtualized based
on different consolidation ratio appropriate for each server class, and again the
carbon footprint will be measured (and non-operational phase carbon-footprint will
be added). The results of the both phases will be then compared to each other to give
an idea of the approximate carbon footprint reduction by using virtualization
techniques.
The definition of different company sizes is different in each country or region;
they also are being referred by different terms across different countries [49].Also,
different companies in different sectors sometimes are measured differently [50]. In
the following chart, companies are divided into five different categories based on
size [51,52,53]:

Table 4.1: Definition of different business sizes based on the number of employees

US

EU

Australia

Small Business

<250

<50

<15

Medium Business

<500

<250

<200

Large Business

<1000 <1000

<500

Enterprise

>1000 >1000

>500

22

Four different companies are assumed each falling into one of the abovementioned
categories:
1. A small business (25 employees)
2. A medium (mid-size) business (175 employees)
3. A large business (750 employees)
4. An enterprise business (3000+ employees)

All of required energy consumption information concerning the physical


hardware used for this study has been collected from analysis results conducted by
Standard Performance Evaluation Corporation (SPEC) which is and provided and
can be found at http://www.spec.org/power_ssj2008/ .

The required HP Servers will be selected based on available best practices and
suggestions about each businesses requirement computing needs (e.g. information
provided in HP Product Bulletin software and HP Power Advisor utility).

MS Excel spreadsheet software has been used for calculation and drawing the
charts.

23

An entry level server (e.g. HP ProLiant ML1xx Series) can serve at least 10
concurrent users. For the small business case study in our research which has 25
users, three servers are chosen to cover its needs.

System 1 The server which has been selected is an HP ProLiant ML110 G5,
equipped with one dual-core 2.66 GHz Intel Xeon Processor 3075 CPU with 4 MB
L2 Cache and 1333 MHz system bus. As can be seen in the test results done by
SPEC, it consumes 97 Watts per hours when 10% utilized and 101 Watts per hour
when being used with 20% CPU utilization. Full results and system details can be
obtained from the link below:

http://www.spec.org/power_ssj2008/results/res2011q1/power_ssj2008-2011012400339.html

Fig. 4.1: Power consumption results of an HP ProLiant ML110 G5

24

A mid-range server (e.g. HP ProLiant DL3xx Series) can serve at least 20


concurrent users. For the medium business case study in our research which has 175
users, nine servers are chosen to cover the companys needs.

System 2 The server which has been selected is an HP ProLiant DL380 G7


equipped with two hexa-core 3.07 GHz Intel Xeon Processor X5675 CPU with 12
MB L3 Cache. As can be seen in the test results done by SPEC, it consumes 93
Watts per hours when 10% utilized and 116 Watts per hour when being used with
30% CPU utilization. Full results and system details can be obtained from the link
below:

http://www.spec.org/power_ssj2008/results/res2011q1/power_ssj2008-2011020900353.html

Fig. 4.2: Power consumption results of an HP ProLiant DL380 G7

25

A high-end server (e.g. HP ProLiant DL5xx Series) can serve at least 40


concurrent users. For the large business case study in our research which has 750
users, twenty servers are chosen to fulfil the companys business requirements.

System 3 The server which has been selected is an HP ProLiant DL580 G5


equipped with four quad-core 1.86 GHz Intel Xeon L7345 CPU with 2x4 MB L2
shared Cache and 1066 MHz system bus. As can be seen in the test results done by
SPEC, it consumes 280 Watts per hours when 10% utilized and 322 Watts per hour
when being used with 40% CPU utilization. Full results and system details can be
obtained from the link below:

http://www.spec.org/power_ssj2008/results/res2007q4/power_ssj2008-2007120700024.html

Fig. 4.3: Power consumption results of an HP ProLiant ML580 G5

26

A carrier-grade-level server (e.g. HP ProLiant Blade Server Series) can serve at


least 50 concurrent users. For the enterprise business case study in our research
which has more than 3000 users, 75 servers are chosen to comply with the
companys business requirements.

System 4 The server which has been selected is an HP ProLiant BL280c G6 Blade
Server equipped with 16 nodes each containing two quad-core 2.27 GHz Intel Xeon
L5520 CPU with 8 MB L2 Cache and 1333 MHz system bus, installed on an HP
BladeSystem c7000 enclosure. As can be seen in the test results done by SPEC, it
consumes 1330 Watts per hours when 10% utilized and 1938 Watts per hour when
being used with 50% CPU utilization. Full results and system details can be obtained
from the link below:

http://www.spec.org/power_ssj2008/results/res2009q3/power_ssj2008-2009063000173.html

Fig. 4.4: Power consumption results of an HP ProLiant BL280 G6

27

According to a study conducted by Greening Greater Toronto in 2010, a vast


number of servers in data centers are operating at four percent utilization on average
basis[54]. Another study by the same firm a year earlier showed the average server
utilization to be between 5 to 7 percent [35]. So, considering 10 percent average
utilization for the whole operational life cycle seems logical. Also, as the Dell Study
considered US grid mix of carbon emission (787 grams CO2eq) for each kWh of
consumed energy (the server in the Dell study used 216 Watts per hour for four years
of non-stop operations, making it consume 7568 kWh of electricity, which resulted
in 5960 of CO2eq carbon emissions), this study will also consider the same amount
of Carbon Emissions for electricity generation. Also, since in the Dell study, 10% of
the total Carbon Footprint accounted for the non-operational stage of the total
cradle-to-grave cycle Carbon Footprint, hence in this study 10 percent extra
Carbon Footprint is calculated and added as well. All the numbers are then
multiplied into the count of servers, based on the number of servers used prior to and
then after Virtualization. The non-operation Carbon Footprint of each server is only
calculated once (prior to Virtualization; to make the results similar to the Dell study)
and then the data is used also for post-Virtualization non-operational stage
calculations.

28

Table 4.2: Pre-virtualization total carbon footprint


Operational
stage
Carbon
Footprint
(CO2eq)

NonOperational
stage
Carbon
Footprint of
each Server
(CO2eq)

Life Stage
Carbon
Footprint
(CO2eq)

Business
Type

Server
Count

Server
Type

Avg.
W
@
10%

Electricity
Consumed
in 4 years
(kWh)

Small

ML110

97

10,197

8,025

267.5

8,827

Medium

DL380

93.6

29,518

23,230

258

25,553

Large

20

DL580

280

196,224

154,428

772

169,871

Ent.

75

BL280c

1330

3,495,240

2,750,754

3,668

3,025,829

Now, following conservative consolidation ratios will be considered for each


business type

Table 4.3: Server consolidation ratios for each business type

Business Type

Consolidation
Ratio

Server Count
Before
Virtualization

Server Count
After
Virtualization

Small

1+1

Medium

Large

20

Enterprise

75

15

29

Note that for the small business, only two servers are virtualized into one server and
one is left untouched (e.g. consolidating less utilized server like DHCP, DNS and
Active Directory Server into one physical server, but leaving the Finance Application
Server which also contains Database Server untouched)

Table 4.4: Post-virtualization total carbon footprint

Power
Usage at
the new
Utilization
Rate

kWh
Used in
Four
Years of
Operation

Life
Stage
Carbon
Footprint
(CO2eq)

Carbon
Footprint
Reduction
(CO2eq) due
to
virtualization

Business
Type

Server
Count

Server
Type

Server
Utilization
due to
Increase
of Load

Small

1+1

ML110
G5

10% +
20%

97 + 101

6,938

5,995

2,832

32

Medium

DL380
G7

30%

116

12,194

10,371

13,210

48.3

Large

DL580
G5

40%

322

56,414

48,258

121,613

71.6

Enterprise

15

BL280c
G6

50%

1938

1,018,613

856,668

2,169,161

71.7

Finally, after servers are virtualized and consolidated into less physical servers, based
on how many servers are consolidated into one physical server, assuming 10%
utilization for each server, the new utilization is calculated based on multiplying the
virtual server count on each physical host into 10% utilization, and then based on the
power consumption ratios provided by SPEC, the new whole life-stage carbon
footprint is again calculated, and then compared to pre-virtualization results. As it
can be seen in the charts, drastic reduction in carbon footprint is shown, ranging
between 32 to 71 percent.

30

180,000.0
160,000.0
140,000.0
120,000.0

Pre Virtualization Life Stage


Carbon Footprint (CO2eq)

100,000.0
80,000.0

Post-Virtualization Life Stage


Carbon Footprint (CO2eq)

60,000.0

Carbon Footprint Reduction


(CO2eq) due to Virtualization

40,000.0
20,000.0
0.0
Small

Medium

Large

Fig. 4.5: Results for Small/Medium/Large Businesses

3,500,000.0
3,000,000.0

3,025,829.0

2,500,000.0

2,169,161.0

2,000,000.0
1,500,000.0
856,668.2

1,000,000.0
500,000.0
0.0
Pre Virtualization Life
Stage Carbon Footprint
(CO2eq)

Post-Virtualization Life
Carbon Footprint
Stage Carbon Footprint Reduction (CO2eq) due to
(CO2eq)
Virtualization

Fig. 4.6: Results for Enterprise Business

31

Carbon Footprint Reduction (%)


Enterprise

Large

Medium

Small
0.0

20.0

40.0

60.0

80.0

Fig. 4.7: Total Carbon Footprint Reduction in Each Business Type

32

Chapter 5

CONCLUSION

By conducting a sample case study on four types of typical business types ranging
from small to enterprise- it was tried to give an approximation of the minimum of
carbon footprint reduction the virtualization technology can bring. As seen in the
case studies, a minimum of 32 percent of total carbon footprint reduction would be
resulted by [partially] implementing and leveraging virtualization into a small
business, and the numbers will drastically improve as the businesses grow and
consolidation ratio increases; in a way that a typical enterprise can reduce its carbon
footprint more than 70 percent. As Virtualization can be implemented in many
different ways, this project certainly cannot study all of implementation methods. But
from this study which can illustrate some typical implementations, it can clearly be
observed how virtualization technology is able to contribute to reducing the carbon
footprint. This study also tried to emphasize on the fact that the power (electricity)
consumption is not the only consequence of computing, and even if all the power is
generated by sustainable/clean energy sources like solar-panel power, wind turbines,
and hydro-electricity, the manufacturing/transportation and then decommissioning
and recycling the computing devices leaves its footprint on the environment. So, as
virtualization will enable businesses to prolong their use of their hardware
investments, not only they will have increased ROI on their IT investments, they will
also help environment by producing less electronics-waste in a certain period of time.
As environmental situation is getting worse on this planet, and the magnitude of the
harm we cause to the nature one of them be the computing- is growing every day,

33

reduction of carbon/environmental footprint to the fullest is the only way to slow the
pace of global warming; and virtualization is undoubtedly one of the best and most
effective answers and solutions to this issues. Further studies need to be done by the
help of hardware manufacturers to better measure and quantify the effectiveness of
virtualization technology in reducing the environmental footprint of computing,
which in turn, push businesses to increase their pace of adopting Virtualization and
Cloud Computing, and motivate governments to enforce more strict rules and
regulations, and through incentives/tax-reduction/etc. motivate more businesses to
shift towards greener computing.

34

REFERENCES

[1] World-o-Meters, real-time world statistics, data provided by Gartner, retrieved on Feb 1st, 2014
from:
http://www.worldometers.info/computers/

[2] How many servers exist in the world, retrieved on Feb 1st, 2014 from:
http://www.quora.com/Servers/How-many-servers-exist-in-the-world

[3] Rackspace will build its own servers just like Facebook and Google do, retrieved on Feb 1st,
2014 from:
http://gigaom.com/2013/01/16/rackspace-will-build-its-own-servers-just-like-facebook-and-googledo/

[4] Assessment of observed changes and responses in natural and managed systems, Rosenzweig et
al. (2007), Cambridge University Press, Cambridge, U.K., and New York, N.Y., U.S.A. pp. 79131

[5] Turn Down the Heat, Why a 4C Warmer World Must be Avoided, November 2012, retrieved on
Feb 1st, 2014 from:
http://climatechange.worldbank.org/sites/default/files/Turn_Down_the_heat_Why_a_4_degree_centri
grade_warmer_world_must_be_avoided.pdf

[6] Interview - Dr Jim Yong Kim, @lliance Magazine, July 2013, retrieved on Feb 1st, 2014 from:
http://www.alliancemagazine.org/node/4312

[7] What's My Carbon Footprint (The Nature Conservancy), retrieved on Feb 1st, 2014 from:
http://www.nature.org/greenliving/carboncalculator

[8] ENERGY STAR, retrieved on Feb 1st, 2014 from:


http://www.energystar.gov/about/

[9] CPU Throttling, Wikipedia, retrieved on Feb 1st, 2014 from:


http://en.wikipedia.org/wiki/CPU_throttling

[10] Virtualization, Wikipedia, retrieved on Feb 1st, 2014 from:


http://en.wikipedia.org/wiki/Virtualisation

[11] Desktop Virtualization, Wikipedia, retrieved on Feb 1st, 2014 from:


http://en.wikipedia.org/wiki/Desktop_virtualization

[12] Our Ecological Footprint Reducing Human Impact on Earth, Mathis Wackernagel & William
Rees (1962) ISBN: 0-86571-312-X

35

[13] Meeting the Carbon Challenge: The Role of Commercial Real Estate Owners, Users &
Managers, Grubb & Ellis Company, Chicago, USA (April 2007), retrieved on Feb 1st, 2014 from:
http://www.grubb-ellis.com/research/Whitepapers/CarbonWhitefnl.pdf

[14] Green Sky Thinking, Patel, J. (2006): Environment Business, issue 122, page 32.
[15] What is Carbon Footprint?, BP (2007) , retrieved on Feb 1st, 2014 from:
http://www.bp.com/liveassets/bp_internet/globalbp/STAGING/global_assets/downloads/A/ABP_AD
V_what_on_earth_is_a_carbon_footprint.pdf

[16] Carbon Footprint of Electricity Generation, POSTnote 268, October 2006, Parliamentary
Office of Science and Technology, London, UK, retrieved Feb 1st, 2014 from:
http://www.parliament.uk/documents/upload/postpn268.pdf

[17] The Carbon Trust Helps UK Businesses Reduce their Environmental Impact, ETAP (2007) ,
retrieved on Feb 1st, 2014 from:
http://www.ec.europa.eu/environment/etap/pdfs/jan07_carbon_trust_initiative.pdf

[18] Ecological Economics Research Trends, NOVA Publishers, 2008, ISBN: 1-60021-941-1

[19] Carbon Footprint of Electricity Generation, POSTnote 383, 2011, Parliamentary Office of
Science and Technology, retrieved on Feb 1st, 2014 from:
http://www.parliament.uk/documents/post/postpn_383-carbon-footprint-electricity-generation.pdf

[20] Green House Gas Inventory Report, United States Environmental Protection Agency, 2013,
retrieved on Feb 1st, 2014 from:
http://www.epa.gov/climatechange/Downloads/ghgemissions/US-GHG-Inventory-2013-MainText.pdf

[21] Electricity Sector Emissions (2011), Climate Change, United States Environmental Protection
Agency, retrieved on Feb 1st, 2014 from:
http://www.epa.gov/climatechange/ghgemissions/sources/electricity.html

[22] Were Green, DreamHost, retrieved on Feb 1st, 2014 from:


http://www.dreamhost.com/about-us/were-green/

[23] Global e-Sustainability Initiative, The Role of ICT in Driving a Sustainable Future, Dec 2012

[24] Uncertainty and Variability in Carbon Footprinting for Electronics - Case Study of an IBM
Rack-mount Server, Christopher Weber, Carnegie Mellon University, December 2010

[25] Carbon Footprint of a Typical Rack Server from Dell, Markus Stutz, November 2011, retrieved
on Feb 1st, 2014 from:

36

http://www.dell.com/learn/us/en/uscorp1/corporate~corp-comm~en/documents~dell-desktop-carbonfootprint-whitepaper.pdf

[26] Energy Consumption of Personal Computing Including Portable Communication Devices, Pavel
Somavat and Vinod Namboodiri (2011) , retrieved on Feb 1st, 2014 from:
http://www.cs.wichita.edu/~vnambood/mypubs/JGEPavel2011.pdf

[27] If Your Appliances Are Avocado, They Probably Arent Green, Alina Tugend, The New York
Times, Business, May 10, 2008, retrieved on Feb 1st, 2014 from:
http://www.nytimes.com/2008/05/10/business/yourmoney/10shortcuts.html?scp=1&sq=appliances%2
0avocado%20green&st=cse

[28] APM BIOS Specification, Microsoft Corporation, Intel Corporation, February 1996, , retrieved
on Feb 1st, 2014 from:
http://download.microsoft.com/download/1/6/1/161ba512-40e2-4cc9-843a923143f3456c/APMV12.rtf

[29] ACPI Overview, ACPI Website, retrieved on Feb 1st, 2014 from:
http://www.acpi.info/presentations/ACPI_Overview.pdf

[30] Kingston Technology, retrieved on Feb 1st, 2014 from:


http://www.kingston.com/dataSheets/KVR16N11_8.pdf

[31] ARM Holdings, retrieved on Feb 1st, 2014 from:


http://www.arm.com

[32] CPU Worlds CPU Comparison Sheet, retrieved on Feb 1st, 2014 from:
http://www.cpu-world.com/info/Intel/Intel_Xeon_7000.html

[33] ARM Cortex-A15 MPCore, retrieved on Feb 1st, 2014 from:


http://www.arm.com/products/processors/cortex-a/cortex-a15.php

[34] HP Case Study, A new day for server hosting, serverCONDO builds its business on HP
Moonshot servers, retrieved on Feb 1st, 2014 from:
http://www8.hp.com/h20195/v2/GetPDF.aspx/4AA4-9986ENW.pdf

[35] Using Virtualization to Improve Data Center Efficiency, The Green Grid, 2009, retrieved on Feb
1st, 2014 from:
http://www.thegreengrid.org/Global/Content/white-papers/Using-Virtualization-to-Improve-DataCenter-Efficiency

[36] An Introduction to Virtualization, retrieved on Feb 1st, 2014 from:


http://www.kernelthread.com/publications/virtualization/

37

[37] Uhlig, R. et al.; "Intel virtualization technology," Computer , vol.38, no.5, pp. 48-56, May 2005

[38] Network Function Virtualization, Introductory White Paper, presented at SDN and OpenFlow
World Congress, October 22-24, 2012, Darmstadt, Germany

[39] Webopedia, Network Virtualization, retrieved on Feb 1st, 2014 from:


http://www.tid.es/es/Documents/NFV_White_PaperV2.pdf

[40] Storage Virtualization Definition, Margaret Rouse, September 2006, Search Storage Magazine,
retrieved on Feb 1st, 2014 from:
http://searchstorage.techtarget.com/definition/storage-virtualization

[41] Webopedia, Storage Virtualization, retrieved on Feb 1st, 2014 from:


http://www.webopedia.com/TERM/S/storage_virtualization.html

[42] Desktop Virtualization Definition, Margaret Rouse, November 2011, Search Virtual Desktop
Magazine, retrieved on Feb 1st, 2014 from:
http://searchvirtualdesktop.techtarget.com/definition/desktop-virtualization

[43] Virtualization Solutions, Types of Virtualizations, Server Virtualization, retrieved on Feb 1st,
2014 from:
http://www.osinet.com/products-a-services/system-integration-and-it-provisioning/virtualizationintegration.html

[44] V-Index: Measuring Virtualization Technology's Reach, July 2011, retrieved on Feb 1st, 2014
from:
http://www.serverwatch.com/trends/article.php/3937621/VIndex-Measuring-VirtualizationTechnologys-Reach.htm

[45] Server Sprawl Definition, Cory Janssen, TechoPedia Online Dictionary, retrieved on Feb 1st,
2014 from:
http://www.techopedia.com/definition/16017/server-sprawl

[46] Wyse Solutions, Green Computing, retrieved on Feb 1st, 2014 from:
http://www.wyse.com/solutions/technologies/green-computing

[47] Desktop Energy Consumption, A Comparison of Thin Clients and PCs, Steve Greenber and
Christa Anderson, prepared for Wyse Technology, retrieved on Feb 1st, 2014 from:
http://www.athena.dk/files/userdir/documents/energy_study.pdf

38

[48] Environmental Benefits of Thin Computing, A Comparison of the Environmental Impacts of


Conventional Desktop and Thin Computing, CanyonSnow Consulting, March 2009, retrieved on Feb
1st, 2014 from:
http://www.wyse.com/sites/default/files/documents/whitepapers/Wyse_Environmental_Benefits_Whit
ePaper.pdf

[49] What is an SME?, European Commission, SME Definition, retrieved on Feb 1st, 2014 from:
ec.europa.eu/enterprise/policies/sme/facts-figures-analysis/sme-definition/

[50] Summary of Size Standards by Industry, U.S. Small Business Administration, retrieved on
Feb 1st, 2014 from:
http://www.sba.gov/content/summary-size-standards-industry

[51] Small Business, Wikipedia, retrieved on Feb 1st, 2014 from:


http://en.wikipedia.org/wiki/Small_business

[52] Small and Medium Enterprises, retrieved on Feb 1st, 2014 from:
http://en.wikipedia.org/wiki/Small_and_medium_enterprises

[53] Guide to Size Standards, retrieved on Feb 1st, 2014 from:


http://www.sba.gov/content/guide-size-standards

[54] Nine Lessons in Greening IT, Greening Greater Toronto, May 2010, retrieved on Feb 1st, 2014
from:
http://www.greeninggreatertoronto.ca/pdf/GGT-Green-Exchange-IT-Summary.pdf

39

S-ar putea să vă placă și