Measuring software architectural stability using retrospective analysis

Abstract

Software architectural stability is concerned with the extent to which an architecture withstands evolutionary changes. An architecture with high stability has a positive impact on the maintenance of software and vice versa. Architectural stability can be viewed from several perspectives and viewpoints. In literature, most of the approaches analyze it from the structural viewpoint based on a host of ambiguous definitions. In this thesis, we disambiguate the definition of stability and adopt the SEI definition of architecture. These definitions are used to identify the factors that affect architectural stability. Based on these factors, we present a survey of the existing architectural sta bility measuring approaches. Also, we use retrospective analysis to develop a new suite of metrics to measure the stability of an architecture from the behavioral viewpoint. A framework was also developed to theoretically validate architectural stability mea sures. The developed suite of stability measures was validated using this framework.