Parallel Changes in Large-Scale Software Development: An Observational Case Study DEWAYNE E. PERRY University of Texas at Austin HARVEY P. SIY Lucent Technologies and LAWRENCE G. VOTTA Motorola, Inc.

1
Parallel Changes in Large-Scale
SoftwareDevelopment An Observational
CaseStudyDEWAYNE E. PERRYUniversity of Texas
at AustinHARVEY P. SIYLucent TechnologiesandLA
WRENCE G. VOTTAMotorola, Inc.

• Greg Holifield

2
Four Essential Problems in Software Development
Brooks 1987

• Evolution
• not only have parallel development within each
  release,
• Among releases as well.
• Scale
• Multiple dimensions of system organization
  1Perry 1996
• Distribution of knowledge

3
Authors Three Goals

• provide a basic understanding of the
  parallel-change phenomena
• begin an investigation of a subproblem
  interfering changes
• explore the relationship between parallel changes
  and the related quality data

4
Methods Toward Each Goal

• basic observational data on the nature of
  parallel changes
• Examine prima facie cases
• changes to changes
• changes within the same day
• summarize available tools that support parallel
  development

5
Classical Tools

• Library-type Systems (check in/out)
• SCCS Rochkind 1975
• RCS Tichy 1982,
• Newer Systems
• Rationals ClearCase(R) Leblang 1994
• Adele Configuration Manager Estublier and
  Casallas 1994

6
Classical Tools

• Integration of Parallel Changes
• permanent versions
• temporary versions
• Semantic Conflicts
• Logical Completeness

7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
Conclusion

• There are multiple levels of parallel
  development.
• The activities within each of these levels cut
  across common files.
• Over the interval of a particular release (I6),
  the number of files changed by multiple MRs is
  60
• though we would expect the degree of awareness
  of the implications of these changes to be higher
  than those made within one day of each other
• There is a significant correlation between files
  with a high degree of parallel development and
  the number of defects