Time-parallel simulation of wireless ad hoc networks with compressed history

Time-parallel simulation (TPS) is a technique which partitions the timespan of the simulation into independently executed simulation segments. Unless the simulated process is regenerative, the output of TPS is only an approximation of the corresponding serial simulation. Previously we have adapted TPS to the simulation of wireless ad hoc networks. By prefixing the measured simulation segment with a warmup interval which can be dynamically extended, we were able to achieve arbitrary accuracy. In general, higher accuracy requires a longer warmup interval, which decreases the speedup. In this paper we introduce compressed history, a technique which improves the performance of TPS for a class of processes which require long warmup intervals to achieve satisfactory accuracy. Compressed history replaces part of the warmup interval, and speeds up the simulation by retaining only those past events that affect significantly the state at the beginning of the measured interval. We present compressed history in general terms, and provide a detailed description of its implementation, tuning, and performance results for the concrete example of the DSDV proactive ad hoc routing protocol.