The paper introduces the concept of model-based performance testing, which we plan to pursue in our research. The underlying idea is to describe various performance aspects as well as functional aspects of a software system using modeling languages like UML, and from the resulting models to automatically design tests that can be used for performance testing. In our research, we also plan to focus on how the modeling and traceability of performance requirements can be achieved across the testing process.

We present an approach for performance testing of software services. We use Probabilistic Timed Automata to model the workload of the system, by describing how different user types interact with the system. We use these models to generate load in real-time and we measure different performance indicators. An in-house developed tool, MBPeT, is used to support our approach. We exemplify with an auction web service case study and show how performance information about the system under test can be collected.

No abstract

No abstract

No abstract