Event scheduling schemes for time warp on distributed systems

In this paper we present event sensitive state saving as a novel idea for sparse state saving in Time Warp synchronized parallel discrete event simulations. This state saving model is aimed at efficient simulation of models where the execution time or granularity for different types or classes of events typically has a large variance. The event sensitive state saving mechanism is sensitive to which class of event the previously executed event belongs, and decide when to save simulation state based on this information. We present an analytical analysis and compare this new state saving model with the performance of an existing adaptive state saving model as well as the traditional Time Warp state saving model with fixed state saving interval.