iDSL: Automated performance evaluation of service-oriented systems

Abstract

Service-oriented systems interconnect with other systems in a time critical manner, making their performance vital. For this purpose, we propose an automated performance evaluation approach for service-oriented systems which includes both performance measurement and prediction. The approach makes use of the iDSL language, a domain specific language tailored to modeling service-oriented systems, and the iDSL toolchain to evaluate iDSL models, as follows. First, discrete-event simulation yields many performance artifacts, e.g., latency breakdown charts, cumulative distribution graphs, and latency bar charts. Second, timed automata-based model checking yields absolute latency bounds. Third, probabilistic timed automata-based model checking leads to exact latency distributions for each service. We successfully validated our approach; several case studies on interventional X-ray systems displayed similar measured and predicted outcomes.