Model-Driven Performance Analysis of UML Design Models Based on Stochastic Process Algebra

The popular model-driven development (MDD) methodology strongly promotes a model-based approach to modular system development, which often implies as an integral part automatic transformation of UML design components into executable forms. When using MDD for verifying performance-related system specifications, UML designs annotated with these specifications in some profile language need to be transformed to stochastic (Markovian) models or timed simulation models. However, most of the previous efforts have focused on transformations of (variants of) UML state machine models and/or transformations to stochastic Petri net models, which lead to two problems: Relying (solely) on state machine models often restricts design flexibility (designers instead prefer choosing diagrams on a case-by-case basis), and graph-oriented Petri net models complicate the modular transformations of UML models. To resolve these problems, we propose stochastic performance analysis of a UML design defined in different sorts of diagrams, including not only state machines but also activity diagrams with temporal annotation in UML-SPT, which are transformed into stochastic process algebraic forms. To our knowledge, this is the first attempt to define stochastic process algebraic semantics for the UML AD with UML-SPT annotations. Unlike the related efforts, ours will facilitate verification in early development stages, in which consultants and architects can benefit from modeling in a natural manner and modular component-based development, thanks to the inherent compositionality of process algebra. Further, to guarantee the validity of the transformation, we have proved the equivalence of our semantics with the stochastic Petri net semantics of UML AD. We have developed a prototype implementation of this performance analysis mechanism, and shown that realistic design examples, defined in different sorts of UML diagrams, can successfully be transformed into those that provide various performance metrics.