Title: A model-driven approach to broaden the detection of software performance antipatterns at runtime

Abstract: Performance antipatterns document bad design patterns that have negative influence on system performance. In our previous work we formalized such antipatterns as logical predicates that predicate on four views: (i) the static view that captures the software elements (e.g. classes, components) and the static relationships among them; (ii) the dynamic view that represents the interaction (e.g. messages) that occurs between the software entities elements to provide the system functionalities; (iii) the deployment view that describes the hardware elements (e.g. processing nodes) and the mapping of the software entities onto the hardware platform; (iv) the performance view that collects specific performance indices. In this paper we present a lightweight infrastructure that is able to detect performance antipatterns at runtime through monitoring. The proposed approach precalculates such predicates and identifies antipatterns whose static, dynamic and deployment sub-predicates are validated by the current system configuration and brings at runtime the verification of performance sub-predicates. The proposed infrastructure leverages model-driven techniques to generate probes for monitoring the performance sub-predicates and detecting antipatterns at runtime.