Development and application of composite complexity models and a relative complexity metric in a software maintenance environment

Abstract

A great deal of effort is now being devoted to the study, analysis, prediction, and minimization of expected software maintenance cost, long before software is delivered to users or customers. It had been estimated that, on the average, the effort spent on software maintenance is as costly as the effort spent on all other software stages. Ways to alleviate software maintenance complexity and high costs should originate in software design. Two aspects of maintenance deserve attention: protocols for locating and rectifying defects and ensuring that no new defects are introduced in the development phase of the software process, and development of protocols for increasing the quality and reducing the costs associated with modification, enhancement, and upgrading of software. This article focuses on the second aspect and puts forward newly developed parsimonious models and a relative complexity metric for complexity measurement of software that were used to rank the modules in the system relative to each other. Significant success was achieved by use of the models and relative metric to identify maintenance-prone modules.