Malaiya and Denton, Module size distribution and defect density, Proc' 11th International Symposium

1
Malaiya and Denton, Module size distribution and
defect density, Proc. 11th International
Symposium on Software Reliability Engineering
(ISRE-00), 2000.
A12

• Geir K. Hanssen

2
Background

• Aims to develop a model to predict defect density
  as a function of module size distribution
• Does the overall defect density of a software
  project vary with its module size distribution?
• And can we exploit this dependence to reduce the
  total number of defects?
• Uses existing data sets to develop/exemplify the
  model
• Two factors that control the defect density at
  release
• the extent and effectiveness of the testing and
  debugging effort
• density present at the beginning of testing
• It is popularly believed that decomposing a
  software system into small modules improves the
  design
• Surprisingly the projects studied show exactly
  the opposite for a large range of module sizes.

3
A composite defect density model

• Two mechanisms that give rise to defects
• module-related are related to how the project is
  partitioned into modules and how the modules
  interact
• instruction-related are associated with the lower
  level building blocks
• In a nutshell
• Defect density
• s size in LOC
• a, b, c are all empirically derived constants a,
  b, c all vary based upon programmer capability
  but are language independent
• Model concept
• a controlled by the defect density of small
  modules
• b controlled by the defect density of small
  modules
• c accounts forthe rise in defect density in
  larger modules
• Optimal module size is
  (results in minimum defect density)

4
Conclusions

• Optimal module size
• 200 LOC?
• Small modules (lt200LOC) should be investigated
  and possibly coalesced
• Observed that for several projects, distribution
  of module sizes is given by an exponential
  expression (unexplained)
• A model for characterizing variation of defect
  density due to module size variation has been
  obtained which can be used as a sub-model for a
  multi-factor defect density model
• Bottom line
• If small modules can be combined into optimal
  sized modules without reducing cohesion
  significantly, than the inherent defect density
  may be significantly reduced