Relaxing causal constraints in PDES

One of the major overheads that prohibits the wide spread deployment of parallel discrete event simulation (PDES) is the need to synchronize the distributed processes in the simulation. Considerable investigations have been conducted to analyze and optimize the two widely used synchronization strategies, namely the conservative and the optimistic simulation paradigms. However, little attention has been focussed on the definition and strictness of causality. Does causality need to be preserved in all types of simulations? Previously, we had suggested an answer to this question. We had argued that significant performance gains can be achieved by reconsidering this definition to decide if the parallel simulation really needs to subscribe to the preservation of causality. In this paper, we investigate this issue even more closely. An in depth analysis using several example simulation models is presented in this paper. In addition, a comparative analysis between unsynchronized and Time Warp simulation is presented.