Design and Analysis of a Minimum Time Buckets Synchronization Algorithm for Parallel and Distributed Simulation

Abstract

Time synchronization is critical for parallel and distributed simulation (PADS). In this paper we first compare the two main synchronization mechanisms in PADS, namely conservative and optimistic protocols. Then we propose a Minimum Time Buckets synchronization algorithm, which can optimistically schedule events without sending anti-messages. Comparing to the well-known Breathing Time Bucket (MTB) synchronization algorithm, our algorithm reduces the waiting time and rollback probability for logical processes, yet it commits the same number of events in each time advance interval. Experiment results show that the MTB algorithm can effectively decrease the rollback rate, reduce the execution time, and achieve a higher speed up.