Performance Comparison of Processor Scheduling Strategies in a Distributed-Memory Multicomputer System

Abstract — Processor scheduling has received con-siderable attention in the context of shared-memory mul-tiprocessor systems but has not received as much atten-tion in distributed-memory multicomputers. Furthermore, most previous research in this area has either used a simu-lation model or an analytical model in evaluating the per-formance of various policies. Very often these models ne-glect several practical aspects of the system and workload. Our goal is to implement processor scheduling policies on a real system and evaluate their performance. We have used a Transputer-based system to implement two policies — one space-sharing policy and one time-sharing policy — in or-der to study the relative performance trade-offs involved. We have also implemented a hybrid space-sharing and time-sharing policy. We report the performance sensitivity of these policies to various system and workload parameters.