Efficient Parallel Simulation of Mobile Wireless Networks by Run-time Prediction of Multi-hop Propagation Delay

In this paper, we propose a method to improve the performance of parallel and distributed simulation of multi-hop mobile wireless networks. We assume that the simulation field is divided into several pieces and is simulated by multiple parallel processes which synchronize conservatively, i.e., without violation of causality. The basic idea of our approach is to predict the lower bounds of timestamps of "potential" upcoming events which cause inter-process messages based on the timestamps of the currently scheduled events plus minimum multi-hop propagation delay estimated in run-time. By this prediction method, we are able to enlarge the lookahead of parallel processes, and their concurrency can be enhanced. We have implemented the parallel and distributed version of our network simulator MobiREAL, and have also implemented this mechanism. The experimental results have shown that we could achieve 50% speed up in average compared with the version using static lookahead in the simulation of end-to-end communications over 1,200 nodes.