A Grid Scheduling Model Based On A Routing-Transferring Method

J Zhang,W Li,Sy Chen,Zw Xu,L Zha

2003-01-01

Abstract:The Grid technology emerges as the need of resource sharing and cooperating in a wide area. Resource scheduling techniques are needed to improve the utilization of these resources. Compared with the traditional computer system, effective resource scheduling in the Grid is difficult because the Grid resources have some new properties: the huge amount, the wide-area distribution, the dynamical changing and being lack of centralized management. In this paper, we propose a Grid resource-scheduling model based on a Routing-Transferring method to solve above problems. Compared with other Grid scheduling models, our model is more measured up to the real Grid environment. The first important feature of our model is that it is a fully decentralized model, which means that there are several independent task queues distributed in the different locations and they are scheduled simultaneously and independently. The second feature is that our model uses the technologies of the resource routing table to tolerate the inaccuracy of the global resource information. We propose a scheduling algorithm called Minimal Q Value (MQV) based on the Routing-Transferring method to schedule a task to a resource as optimally as possible. The kernel of the MQV algorithm is that it will schedule a task on a nearest resource. Only when the nearer resource is overload, the MQV algorithm schedules the task to a farther resource. We implement our model on a Grid simulation tool called SimGrid The result of the simulation shows that our model is efficient and more suitable for the Grid-level resource scheduling.