System structure and software maintenance performance

An experiment is designed to investigate the relationship between system structure and maintainability. An old, ill-structured system is improved in two sequential stages, yielding three system versions for the study. The primary objectives of the research are to determine how or whether the differences in the systems influence maintenance performance; whether the differences are discernible to programmers; and whether the differences are measurable. Experienced programmers perform a portfolio of maintenance tasks on the systems. Results indicate that system improvements lead to decreased total maintenance time and decreased frequency of ripple effect errors. This suggests that improving old systems may be worthwhile and may yield benefits over the remaining life of the system. System differences are not discernible to programmers; apparently programmers are unable to separate the complexity of the systems from the complexity of the maintenance tasks. This finding suggests a need for further research on the efficacy of subjectively based software metrics. Finally, results indicate that a selected set of automatable, objective complexity metrics reflected both the improvements in the system and programmer maintenance performance. These metrics appear to offer potential as project management tools.