An Analytical Comparison between Pair Development and Software Development with Inspection By:

advertisement
An Analytical Comparison between
Pair Development and Software
Development with Inspection
By:
Monvorath (Molly) Phongpaibul
phongpai@usc.edu
Present to:
USC- CSSE- ARR 2007
February 13, 2007
2/13/07
(c) USC-CSSE
1
Content
•
•
•
•
•
2/13/07
Research Questions
Research Methodology
Quantitative Results
Conclusion
Future Work
(c) USC-CSSE
2
Research Question
• There are three main research
questions:
– What are the commonalities /
differences and relative strengths
/ weaknesses of each practice?
– Under what conditions you might
prefer one over the other?
– Under what conditions you would
merge the two practices?
2/13/07
(c) USC-CSSE
3
Content
•
•
•
•
•
2/13/07
Research Objective and Approach
Research Methodology
Quantitative Results
Conclusion
Future Work
(c) USC-CSSE
4
Pair Development (PD) Process
2/13/07
(c) USC-CSSE
5
Software Cost of Quality (CoSQ)
[Krasner, 1998]
TDC: Total Development Cost
C production  CTDC  C quality
C quality  C prevention  C appraisal  C rework
2/13/07
(c) USC-CSSE
6
Activities of CoSQ
Conformance Costs
Prevention Costs








Prototyping
User requirement
reviews
Quality planning
Training
Reuse library
Process
improvements
Metrics collection
and analysis
Quality standards
2/13/07
Non-Conformance Costs
Appraisal Costs







Inspection / peer
review
Continuous review
Testing
Software quality
assurance
V&V activities
Quality audits
Field performance
trails
(c) USC-CSSE
Rework Costs






Fixing defects
Corrective rework
Re-inspection, re-review,
re-testing
Re-design
Updating documents
Integration
7
Content
•
•
•
•
•
2/13/07
Research Objective and Approach
Research Methodology
Quantitative Results
Conclusion
Future Work
(c) USC-CSSE
8
TDC & CoSQ Results
TDC
(man-hour)
Production
Costs
(man-hour)
Appraisal
Costs
(man-hour)
Rework
Costs
(man-hour)
E1
(Thailand 05)
PD Group
526.73
314.02
102.07
8.03
FI Group
695.11
309.23
234.97
43.72
E2
(Thailand 05)
PD Group
336.66
186.67
73.33
13.67
FI Group
482.5
208.5
165
45
E3
(Thailand 05)
PD Group
1392.9
654.2
325.7
233
FI Group
1342
429
436
317
E4
(US 05)
PD Group
187.54
68.16
88.83
20.05
FI Group
237.93
62.82
122.10
42.52
2/13/07
(c) USC-CSSE
9
Costs VS. Quality
Team
TDC
#Test Defects
E1
(Thailand 05)
PD Group
526.73
4.429
FI Group
695.11
5.142
E2
(Thailand 05)
PD Group
336.66
0
FI Group
482.5
0
E3
(Thailand 05)
PD Group
1392.9
21 (11 Major)
FI Group
1342
29 (18 Major)
E4
(US 06)
PD Group
187.54
6.8 and 1.4
FI Group
237.93
6.5 and 2.0
2/13/07
(c) USC-CSSE
10
Costs by Phase (E4)
Team#
PD
Group
FI
Group
2/13/07
Req.
Des.
Imp.
Test
P1
P2
P3
16.05
13.95
16.20
41.75
41.42
35.70
63.73
64.37
66.53
26.65
25.33
27.12
P4
P5
I1
I2
16.77
12.40
26.83
22.65
34.30
37.92
40.78
41.50
67.03
73.03
96.33
90.27
27.68
32.05
21.33
35.30
I3
I4
23.25
21.70
38.00
39.37
92.30
84.83
31.40
43.98
(c) USC-CSSE
11
Defect Type Analysis
• Test defects were classified by
Orthogonal Defect Classification
(ODC) v.5.11
• For requirement defects, pair development is
better on detecting ambiguity/testability and
correctness defects. Inspection is better on
detecting consistency defects
• For Design / Code defects, the weakness of pair
development is internal and external interfaces.
2/13/07
(c) USC-CSSE
12
Content
•
•
•
•
•
•
•
2/13/07
Research Objective and Approach
Research Background
Research Methodology
Quantitative Results
Qualitative Results
Conclusion
Future Work
(c) USC-CSSE
13
Conclusion
• PD offers an option of reducing the
schedule.
• PD spent less total development cost
than inspection with the same level
of quality.
– Due to less appraisal costs and failure
costs
• Experiment provided insights on
when to use pair development or
inspections
2/13/07
(c) USC-CSSE
14
Example of Pair Development
Scenario
Team Size
5
0
5
5
Time to Market
Safety Criticality
2/13/07
(c) USC-CSSE
15
Example of Software Development
with Inspection Scenario
Team Size
5
0
5
5
Time to Market
Safety Criticality
2/13/07
(c) USC-CSSE
16
Example of Combination
between Both Practices Scenario
Team Size
5
0
5
5
Safety Criticality
2/13/07
Time to Market
(c) USC-CSSE
17
Content
•
•
•
•
•
•
•
2/13/07
Research Objective and Approach
Research Background
Research Methodology
Quantitative Results
Qualitative Results
Conclusion
Future Work
(c) USC-CSSE
18
Future Work
• Simulate the Pair
Development Model.
– Develop an Extension to
Ray Madachy’s System
Dynamics model of
inspections, to compare pair
development and inspection
dynamics and calibrate it to
the experimental data.
– Software development
spending profiles analysis.
2/13/07
(c) USC-CSSE
19
Questions & Answer
– Thank You –
2/13/07
(c) USC-CSSE
20
Download