Performance Analysis of Legacy Perl Software via Batch and Interactive Trace Visualization
Performing an analysis of established software usually is challenging. Based on reverse engineering through dynamic analysis, it is possible to perform a software performance analysis, in order to detect performance bottlenecks or issues. This process is often divided into two consecutive tasks. The first task concerns the monitoring of the legacy software, and the second task covers analysing and visualizing the results. Dynamic analysis is usually addressed via trace visualization, but finding an appropriate representation for a specific issue still remains a great challenge. In this paper we report on our performance analysis of the Perl-based open repository software EPrints, which has now been continuously developed for more than fifteen years. We analyse and evaluate the software using the Kieker monitoring framework, and apply and combine two types of visualization tools, namely Graphviz and Gephi. More precisely, we employ Kieker to reconstruct architectural models from recorded monitoring data, based on dynamic analysis, and Graphviz respectively Gephi for further analysis and visualization of our monitoring results. We acquired knowledge of the software through our instrumentation and analysis via Kieker and the combined visualization of the two aforementioned tools. This allowed us, in collaboration with the EPrints development team, to reverse engineer their software EPrints, to give new and unexpected insights, and to detect potential bottlenecks.