Performance is a critical non-functional parameter for real-time systems and performance analysis is an important task making it more challenging for complex real-time systems. Mostly performance analysis is performed after the system development but an early stage analysis and validation of performance using system models can improve the system quality. In this paper, we present an early stage automated performance evaluation methodology to analyse system performance using the UML sequence diagram model annotated with modeling and analysis of real-time and embedded systems (MARTE) profile. MARTE offers a performance domain sub-profile that is used for representing real-time system properties essential for performance evaluation. In this paper, a transformation technique and transformation rules are proposed to map the UML sequence diagram model into a Generalized Stochastic Timed Petri net model. All the transformation rules are implemented using a metamodel based approach and Atlas Transformation Language (ATL). A case study from the manufacturing domain a Kanban system is used for validating the proposed technique.
Read full abstract