Documente Academic
Documente Profesional
Documente Cultură
Necessarily Complex
-Himanshu Pandey
Nishu Lohia
(Aricent Technologies)
Process Performance Models - Not Necessary Complex
HMBP 2010
21-May-2010
Who We Are
Equipment Manufacturers
Device Manufacturers
Service Providers
7-Jun-10 3
What We Offer
7-Jun-10 4
Agenda
Problem Statement
Aricent’s way to Resolution
Process Performance Modelling – Overview
Models- Overview
– Rayleigh’s Defect Prediction Model
– Test End date prediction Model
– NHPP + Gompertz Model
Combinations of Models
Conclusion
Q&A
7-Jun-10 5
Problem Statement
7-Jun-10 6
Aricent’s Way Of Resolution
• Identify generic life cycle steps where PPM could be applied across
maximum projects
• Identify in-house developed statistical tools which are being used most
frequently across the organization.
• Identify the tools those can be used in collaborative mode effectively to:
• Perform What if Analysis
• Implement PDCA cycle
7-Jun-10 7
Process Performance Modeling
7-Jun-10 8
In-house Statistical Tools
Rayleigh’s Defect Prediction Model
Test End Date Prediction Model
NHPP + Gompertz Model
7-Jun-10 9
In-house developed Models Overview
7-Jun-10 10
Rayleigh’s Defect Prediction Model
• Benefits:
120
Cum_Defects
60
100
50
80
40
60
10 20
7-Jun-10 11
In-house developed Models Overview
7-Jun-10 12
Test End Date Prediction Model – Overview
40.00
35.00
30.00
25.00
20.00
15.00
10.00
5.00
0.00
1 2 3 4 5 6
7-Jun-10 13
Test End Date Prediction Model cont..
• Benefits
• Strategizing remaining number of days in order to finish testing on time
Replicate1
350 100%
Replicate2
X Axis - Days to finish testing
Y Axis - Frequency
Replicate3
90%
Replicate4
300
Replicate5
80%
Replicate6
Replicate8
200
60% Replicate9
Number of
days
Replicate10
50% Replicate11
150 Replicate12
40%
Replicate13
Replicate14
100 30%
Replicate15
Replicate16
20%
Replicate17
50
10% Replicate18
Replicate19
0 0% Replicate20
14.7 to 14.75
14.8 to 14.85
14.9 to 14.95
15 to 15.04
15.09 to 15.14
15.19 to 15.24
15.29 to 15.34
15.39 to 15.44
15.49 to 15.54
15.59 to 15.64
15.69 to 15.74
15.79 to 15.83
15.88 to 15.93
15.98 to 16.03
16.08 to 16.13
16.18 to 16.23
16.28 to 16.33
16.38 to 16.43
16.48 to 16.53
16.58 to 16.62
16.67 to 16.72
16.77 to 16.82
16.87 to 16.92
16.97 to 17.02
17.07 to 17.12
17.17 to 17.22
17.27 to 17.32
17.37 to 17.41
17.46 to 17.51
17.56 to 17.61
% Distribution of
Replicate1
90%_Line
7-Jun-10 14
In-house developed Models Overview
7-Jun-10 15
NHPP + Gompertz Model – Overview
CURRENT STATUS OF
• Performs forecast on SOFTWARE RELIABILITY
DATE FOR WHICH S/W RELIABILITY
• Reliability of the software under testing
23-Mar-10
STATUS IS GIVEN
SOFTWARE RELIABILITY ON ABOVE
79.72%
MENTIONED DATE
7-Jun-10 16
NHPP + Gompertz Model Cont..
• Benefits:
• Strategizing remaining number of testing days in order to improve reliability and minimize
failures passing to customer
7-Jun-10 17
Process Performance Models
NHPP- Gompertz + Test End Date Prediction
Rayleigh’s Model + Test End Date Prediction
NHPP-Gompertz Model + Test End Date
Prediction Model+ Rayleigh’s Model
7-Jun-10 18
NHPP- Gompertz + Test End Date Prediction
• Test End Date Prediction Tool takes test execution data as input and provides the number of days
pending to finish the testing
• NHPP Model, along with execution data, require to know how many more days testing will
continue in order to predict the failures
• Number of failed test cases predicted by NHPP then in turn act as revised input for Test End Date
Prediction to re-calibrate the testing end date
• Perform multiple calibrations in order to arrive at best suitable situation to strategize where
highest software reliability can be achieved within lesser time and lesser failures
7-Jun-10 19
Rayleigh’s Model + Test End Date Prediction
• Test End Date Prediction Tool takes test execution data as input and provides the number of days
pending to finish the testing
• Rayleigh’s Model, require to know for how many days testing will continue in order to predict the
defects
• Derive number of test case failures from number of defects predicted by Rayleigh’s model using
baseline defects/failure rate and use this as input in Test End Date prediction for revised number
of days.
• Perform multiple calibrations in order to know estimated defect leakage to customer and
strategize for the prevention.
7-Jun-10 20
NHPP-Gompertz + Test End Date Prediction + Rayleigh’s Model
• Test End Date Prediction Tool usage along with NHPP and Rayleigh models separately has been
discussed
• Now, comparing the results of two, brings down to an interesting analysis in order to verify the
defects/TC failure rate with baseline figures.
• Hence, knowing your current project’s performance in comparison to past project’s performance
justifies our subjective confidence on how far/close we are w.r.t. to achieving our targets.
• Re-calibration and verification analysis based PCDA format, helps in strengthening our prevention
strategies and hence achieving great results.
7-Jun-10 21
NHPP-Gompertz + Test End Date Prediction + Rayleigh’s Model
• Special scenario:
• When Rayleigh’s and NHPP’s forecasts are not in synch:
• Study the phase wise defect distribution curve
• Study Actual value curve vs Estimated value curve
• Identify and Analyze the phases outside our control
• Study the co-efficient of determination
• Correctness of available data
• Availability of support data used for forecasts
• Decision
• Choose the model for which most of the above parameters are satisfied
7-Jun-10 22
Conclusion
• Yes! We applied these in-house developed PPMs and achieved CMMI level 5 (v1.2) successfully.
• Applied on wide range of Projects.
• The models applied ,undoubtedly:
• Has almost no Cost
• Are easy to understand and implement
• Are Easily Customizable
• Are not Overhead to Implement
• Implies no resistance to use
• Has no constraints of availability
• Satisfies CMMI practices completely
7-Jun-10 23
Thank You!
7-Jun-10 24
Click here for:
Click here