A Compositional Semantics for UML State Machines Aimed at Performance Evaluation

Unified Modeling Language (UML) is gaining acceptance to describe the behaviour of systems. It has attracted the attention of researchers that are interested in deriving, automatically, performance evaluation models from system’s descriptions. A required step to automatically produce a performance model (as any executable model) is that the semantics of the description language is formally defined. Among the UML diagrams, we concentrate on States Machines (SMs) and we build a semantics for a significant subset of them in terms of Generalized Stochastic Petri Nets (GSPNs). The paper shows how to derive an executable GSPN model from a description of a system, expressed as a set of SMs. The semantics is compositional since the executable GSPN model is obtained by composing, using standard Petri net operators, the GSPN models of the single SMs, and each GSPN model is obtained by composition of submodels for SM basic features.