The Use of Product Measures In Planning Monitoring and Controlling Software Maintenance Releases

• Main Author: Ware, Melanie
• Published: Ulster University 2008
• Subjects: 005.3
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.490390

This work concerns product measures, in the management of software maintenance releases. It describes the implementation of a measurement program, relating to a commercial application, written in C++ and supplied by Meridio. Three levels of analysis are applied and their effectiveness, to inform on the product, determined. Their potential use, in relation to quality and process models, and the commercial context, is assessed. The measurement programme uses product measures, from the Chidamber and Kemerer metric suite, supplemented by measures of class service utilisation. A codeanalyser is used to extract data-sets from the application. The practicalities related to this approach are discussed. Initially, the application was surveyed and each release characterised. This contributed to the formulation of an evolutionary profile, of the application, which was contrasted to findings, relating to existing studies. A maintenance, change-prone, release was subjected to further study. The ability to predict change-prone classes was determined, using Pareto analysis. Calibration, of the selection criteria, was examined in relation to predictive performance. The capacity to predict classes, with a high change-density, or a high change-volume, was studied. With regard to the prediction of change-prone classes, the outcome, in relation to these classes and future product refactoring effort, is related to refactoring tactics. Phase analysis techniques are applied, to the application, on a monthly basis, over two years and three releases. Four analyses are developed, concerning the degree and nature of change, at class and application level. The ability of the analyses, to detect change in the emerging release, is contrasted to information from the project team, to documentation and to file modification counts. The analyses, successfully, reflect change. The release cycles, of three commercial organisations, are profiled. It is clear, that, it would be beneficial to these organisations, if they could use the proposed measures to monitor for change.