The performance of adaptive protocols using either aggressiveness control, risk control or both are compared to each other and an optimistic protocol. Empirical results are presented from execution on a CM-5 using up to 256 nodes. The results demonstrate that between the aggressiveness and risk control protocols, the aggressiveness control protocol performs better overall; however, risk control performs best in some situations. A protocol combining aggressiveness and risk control is more consistent at performing better than an optimistic protocol. Also, the use of straggler information is seen to enhance the risk control protocol.
Keywords: Parallel and Distributed Discrete Event Simulation, Adaptive Protocol, Time Warp, Closed Queueing Systems.