MBPeT: A model-based performance testing tool

In recent years, cloud computing has become increasingly common. Verifying that applications deployed in the cloud meet their performance requirements is not simple. There are three different techniques for performance evaluation: analytical modeling, simulation, and measurement. While analytical modeling and simulation are good techniques for getting an early performance estimation, they rely on an abstract representation of the system and leave out details related for instance to the system configuration. Such details are problematic to model or simulate, however they can be the source of the bottlenecks in the deployed system. In this paper, we present a model-based performance testing tool that measures the performance on web applications and services using the measurement technique. The tool uses models to generate workload which is then applied to the system in realtime and it measures different performance indicators. The models are defined using probabilistic timed automata and they describe how different user types interact with the system. We describe how load is generated from the models and the features of the tool. The utility of the tool is demonstrated by applying to a WebDav case study.