Scalable Scheduling on a Network of Workstations.

Sanglu Lu,Li Xie

DOI: https://doi.org/10.1145/346152.346330

2000-01-01

Abstract:This paper presents an adaptive scalable scheduling model ASCS based on NOWs, which provides a dynamic load sharing facility by allowing tasks to scale up to utilize desired excess computing capacity depending upon available workstations on a NOW, and also scale down to maintain workstation autonomy. In this paper, a hierarchical scheduling policy is described: the inter-cluster scheduling provides a simple, fast and less-overhead load sharing among clusters in the system; and the intra-cluster scheduling provides a high performance scheduling in a cluster, and an adaptive coscheduling is implemented, some scalable algorithms are introduced to schedule those parallel/distributed applications to adapt to system changes. The simulation results show that the ASCS Model has better performance with less overhead, and it can well adapt to different system structures, and also well adapt to large scale systems.