Using Software Metrics Burak Turhan

1
Using Software MetricsBurak Turhan
2
Outline

• Introduction
• HP Experience
• NASA Experience
• Relationships between concepts (i.e. SC
  Complexity Maintanence diffculty).
• Conclusions

3
Papers

• (IEEE Computer Special Issue on Software Metrics,
  1994)
• R.Grady, Successfully Applying Software Metrics
• D.L.Lanning and T.M.Khoshgoftaar, Modeling the
  Relationship Between Source Code Complexity and
  Maintenance Difficulty
• G.Stark, R.Durst and C.W.Vowell, Using Metrics
  in Management Decision Making
• C.Jones, Software metrics

4
Introduction

• Success is marked by a measurable end point.
• Software metrics are measures of success for
  software products and development processes.
• Many sophisticated tools with many possible
  project variables. Keep it simple.

5
Introduction

• Remember the real meaning of using metrics and
  models.
• Create dialogue btw. Managers and developers or
  customers and suppliers.
• Goal of prediction is to compare the actual and
  predicted.
• Observation Useful results are generally
  obtained from simple models rather than
  intractable and non-interpretable exact, complex
  models.
• When the results are not reasonable, actions
  should be taken.
• Be careful while using metrics.

6
Major Uses of Software Metrics
7
Project Estimation and Progress Monitoring

• The basis for estimates ? Data collection.
• The bottom line for high level managers Is the
  project on schedule?
• Schedule slip.
• Monitoring progress against estimates.
• Tracking functionality
• Found and fixed defects

8
Schedule Slip
9
Tracking Functionality
10
Found and Fixed Defects
11
Evaluation of Work Products

• Work product
• An intermediate or final output that
  describes the design, operation, manufacture or
  test of some portion of a deliverable or salable
  product. It is not the final product.
• Strong bias to code metrics in metric
  applications, due to ease of extraction.
• Cyclometric complexity.
• Fanout for data-oriented/design complexity

12
Properties of Metrics

• NASA MOD initiated software metrics program in
  May 1990
• Relevant to the environment.
• Cost-effective collection and analysis.
• Multiple metrics to crosscheck.
• Strong basis in industry practice for setting
  rule-of-thumb thresholds.
• 3-step process to implement
• Definition
• Documentation
• Education

13
Problem 1

• How to tell if the project is on schedule and
  within budget?
• Earned-value charts.

14
Problem 2

• How hard will it be for another organization to
  maintain this software?
• McCabe Complexity

15
Problem 3

• How to tell when the subsystems are ready to be
  integrated
• Defect Density Metrics.

16
Problem 4

• How to tell if a test schedule makes sense?
• Historical data.

17
Canonical Correlation Analysis
18
Preliminary Work
19
Conclusions

• Rather than seeking for new metrics, seek for
  efficent use of existing hundreds of metrics.
• Build simple models that are easily
  interpretable.
• Analyse several types of relationships. (i.e.
  Linear, non-linear, observable latent
  variables, combination of models in serial or
  parallel).

20
Thanks !

• Questions ?