Parallel CSRL Model Checking: First Results and Pointers to the Future

The verification of quantitative aspects of a system, like performance and dependability, by means of model checking has become an important and vivid area of research over the past decade. Recently, great progress has been made in the combined analysis of a system's performance and dependability (the so-called performability) by means of model checking. Both the logic CSRL (continuous stochastic reward logic) and a number of model checking algorithms for this logic are tangible outcomes of this research. To evaluate CSRL properties, however, it is necessary to solve large systems of partial differential equations (PDEs). The inherent time complexity of the model checking algorithms makes CSRL model checking for system models with more than 100,000 states practically infeasible. To overcome these difficulties, we have investigated various possibilities to parallelise the CSRL model checking algorithms. First practical experiments with a parallel version of the so-called path-exploration-based algorithm will be reported.