Scalable Scheduling on a Network of Workstations*
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 1simulation 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.
版权说明：以下全部内容由陆桑璐上传于 2005年05月11日 19时31分17秒，版权归本人所有。