Distributed Branching Bisimulation Reduction of State Spaces

AbstractEnumerative model checking tools are limited by the size of the state space to which they can be applied. Reduction modulo branching bisimulation usually results in a much smaller state space and therefore enables model checking of much larger state spaces. We present an algorithm for reducing state spaces modulo branching bisimulation which is suitable for distributed implementation. The target architecture is a cluster with a high bandwidth interconnect. The algorithm is based on partition refinement and it works on transition systems which contain cycles of invisible steps, without eliminating strongly connected components first. To avoid fine grained parallelism, the algorithm refines the whole partition instead of just a single block in the partition. We prove correctness and also present some experimental results obtained with single threaded and distributed prototypes