David Parker
Reader, Computer Science, University of Birmingham
[KPZM04] Marta Kwiatkowska, David Parker, Yi Zhang and Rashid Mehmood. Dual-Processor Parallelisation of Symbolic Probabilistic Model Checking. In Proc. 12th International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS'04), pages 123-130, IEEE CS Press. October 2004. [ps.gz] [pdf] [bib]
Downloads:  ps.gz ps.gz (59 KB)  pdf pdf (154 KB)  bib bib
Abstract. In this paper, we describe the dual-processor parallelisation of a symbolic (BDD-based) implementation of probabilistic model checking. We use multi-terminal BDDs, which allow a compact representation of large, structured Markov chains. We show that they also provide a convenient block decomposition of the Markov chain which we use to implement a parallelised version of the Gauss-Seidel iterative method. We provide experimental results on a range of case studies to illustrate the effectiveness of the technique, observing an average speed-up of 1:8 with two processors. Furthermore, we present an optimisation for our method based on preconditioning.