Applying and Adapting a Performance Deviation Analysis Tool in Web Information Systems

PerfMiner is a tool that allows the identification of performance deviations during the evolution of software systems. This paper describes the experience of applying and adaptating the PerfMiner tool to the performance analysis of large-scale web information systems. The main result of this experience is the development of the Visum tool that allows detecting, analyzing and comparing performance deviations that occur in testing and production environments of a system. We also report performance issues of 4 different categories found in the analyzed systems, as well as specific refactorings that can help to address them. Finally, we also present discussions and lessons learned related to the automation of performance evaluation activities using our tools.