Modular Performance Modelling of Mobile Applications using Graph Transformation

Graph transformation provides a visual and formal notation for modelling systems of dynamic nature. We use graph transformation for modelling mobility and performance, and provide a methodology for modular system modelling to handle scalbility issues of large systems. In our methodology we have distinguished three approaches for system modelling, monolithic, topdown and bottom-up. In the monolithic approach, a system is modelled as a global or whole-world system. In the top-down approach, a global system is projected to its views based on their local type graphs. In the bottom-up approach, a system is modelled as a set of subsystems with shared interface. A whole system is composed from its subsystems. We generate labelled transition systems (LTSs) from graph transformation systems/views in GROOVE and transform them into Continuous Time Markov Chains (CTMCs). These CTMCs are further translated into the Performance Evalution Process Algebra (PEPA) or PRISM. In PEPA and PRISM subsystems are synchronized over shared labels to compose a global system. We demonstrate that the composed model is bisimilar to its original global model. In addition stochastic analysis of models are also carried out in PEPA or PRISM for performance checking. We have given tool support for view generation from a global graph transformation system in GROOVE and transforming LTSs generated from graph transformation systems, into CTMCs, and CTMCs into PEPA or PRISM models.