An approach to scheduling parallel applications on networks of workstations

Networks of Workstations (NOWs) are increasingly being used as platforms for running distributed applications. One of the fundamental problems that must be addressed to realize good performance for distributed applications on NOWs is the determination of no efficient job scheduling algorithm. This paper presents a highly scalable algorithm for scheduling distributed applications on NOWs. The performance of the proposed job scheduling algorithm is compared with a space-sharing scheduling policy. Our results show that the proposed scheduling policy performs substantially better than the space-sharing policy.