Time-Parallel Simulation for Stochastic Automata Networks and Stochastic Process Algebra.

Time Parallel Simulation (TPS) is the construction of the time-slices of a sample-path on a set of parallel processors (see [11] chap. 6 and references therein). TPS has a potential to massive parallelism as the number of logical processes is only limited by the number of time intervals which is a direct consequence of the time granularity and the simulation length. Stochastic Automata Networks (SAN in the following) and some stochastic process algebra (like PEPA) allow the construction of extremely large Markov chains which are difficult to analyze due to their size. Here, we show how we can use TPS to solve efficiently some models based on SAN or PEPA. The approach uses some graph theoretical properties which can be checked easily on a SAN or a PEPA model. The quantitative results are obtained by a TPS based on linear recurrence equations of the daters with associative operators.