Abstract: We develop a general model for software
development process and
propose a policy to manage system coordination using system fault
reports (e.g., interface inconsistencies, parameter mismatches,
etc.). These reports are used to determine the timing of
coordination activities that remove faults. We show that under an
optimal policy, coordination should be performed only if a "threshold" fault count has been exceeded. We apply the policy to
software development processes and compare the management of those
projects under different development conditions. A series of
numerical experiments are conducted to demonstrate how the fault
threshold policy needs to be adjusted to changes in system
complexity, team skill, development environment, and project
schedule. Moreover, we compare the optimal fault threshold policy to
an optimal release-based policy. The release-based policy does not
take into account fault data and is easier to administer. The
comparisons help to define the range of project parameters for which
observing fault data can provide significant benefits for managing a
software project.