SOFTWARE METRICS

1
SOFTWARE METRICS

• MN3309 Session 27
• Prof. Mark Nissen

2
Agenda

• Web Software Example
• Mythical Manmonth
• Software Metrics
• Metrics Exercises
• Summary

3
Web Software Example

• Web site has many pages (IOC)
• New addition (FOC)
• Create using word processor, ftp to site
• 10 pages (static) - 10 IF, 10 EQ
• 1 image (x10) - 10 IF
• 10 links to external sites (x10) - 100 EQ
• 10 links from IOC app (x10) - 0 EIF
• Consistency
• Important to be consistent in counting
• Ensure common interpretation/measure

4
Counting Example
(0) (0) (10)
(0) (0) (0) (100)
(20)
5
Counting Example
6
Calibration

• Past project
• Same environment, counting rules
• 25 AFP/page
• 4 SLOC/AFP
• 320 hours for 15 new pages
• Parametrics
• .85 hours/AFP (15 x 25)1.0 (productivity? scale?)
• Ec A(KSLOC) --gt A Ec/KSLOC 213
• Ep 213 x 1.0 213 hours
• Reasonable? Concerns?

7
Mythical Manmonth

• Programmers are optimists?
• Manmonths mythical?
• 2 components?
• 30/year max manpower buildup
• ROT for scheduling
• Coding x?
• Planning y? Test integration z
• Gutless estimating?
• Sharp milestones?

8
Estimate Bootstrapping

• Perform a sample of work, record time
• Time by task types (plan, meet, code, etc.)
• Use to make calibrate estimates
• 1 new Web page
• 2 hours to create (multiply by 6?)
• How long to plan? Test? Integrate?
• Time associated with old Web pages?
• Time not associated with Web pages?
• Triangulate with other methods

9
Software Measurement

• Measurement Motivation
• Measurement Life Cycle
• Metrics Usage Groundrules
• Examples of Metrics
• Cautions About Metrics

10
Measurement Motivation

• Key to process improvement
• cant manage what you cant measure
• Monitor risk areas, before crisis
• Basis for rewards incentives
• Key how to measure tech progress?
• C-SAWS?

11
Measurement Life Cycle
12
Metrics Usage Groundrules

• Metrics must be
• Understandable?
• Economical?
• Field tested
• Highly-leveraged
• Timely
• Improvement-oriented
• Applied to all life cycle phases
• Useful at multiple levels
• Metrics related to estimates?

13
Typical Software Metrics

• Quality - user satisfaction, Rome Labs
• Size - SLOC, function/feature points
• Complexity - McCabe, Halstead
• Requirements - Stability, traceable
• Effort productivity - mm/SLOC or FP
• Cost schedule - /mm mo, phased
• Scrap rework - defect/correct rates
• Support - track characteristics (size)

14
Army STEP Metrics

• Schedule cost
• Computer resource utilization
• Contractor SEE rating
• Design requirements stability
• Fault profile
• Complexity
• Breadth depth of testing
• Reliability

15
F-22 Time-Phased Metrics
16
Program Feedback Control
17
Cautions About Metrics

• Use as indicators, not absolutes
• Only as good as underlying data
• CDRL items program tracking
• Do not measure everything
• Some metrics universal
• Many program-idiosyncratic
• Evolve with program
• Use multiple metrics, track estimates
• Tie metrics to risk areas problems

18
Change Point Tracking
19
ERROR SCRs-DISCOVERY CORRECTION
CURRENT STATUS
R
A key positive indicator will be when the
discovery line flattens out as the rate of new
errors encountered approaches zero.
The correction rate for Priority 1 2 SCRs
continues to keep pace with the discovery rate.
20
Program Stretch-out Effects 1
Baseline Program
P P-Mo, Mo B/L 140, 5
C
T
D
A
I
21
Program Stretch-out Effects 2
Baseline Program
P P-Mo, Mo B/L 140, 5 S-P 140, 7
plan
Stretch-out
22
Program Stretch-out Effects 3
Baseline Program
P P-Mo, Mo B/L 140, 5 S-P 140, 7 S-A 215, 8
plan
actual
Core staff
Stretch-out
23
Summary

• Measurement is important
• Key to performance improvement
• ID problems in advance of crisis
• Metrics must be tailored
• Some metrics universal
• Many program-idiosyncratic
• Evolve with program