Research On Study And Implementation Of Grid Resource Management And Scheduling Algorithm

Grid computing, as a key technology for the next generation Internet, becomes one of the hot areas of current research at domestic and international. Resource management and scheduling algorithm are the two fundamental problems in grid computing. Based on the platform of the main stream grid middleware Globus, this article focuses on the resource management and scheduling algorithm of grid computing in the context of its application in campus computing grid.This article first analyzes a paradoxical phenomenon that is predominant in campus LAN settings where there is a large amount of idle CPU cycles available while at the same time the computing needs from many researchers can't be met.It proposes an idea of setting up a campus computing grid with a detailed plan for its implementation,discusses the challenges of upscaling a campus computing grid into an Internet computing grid and suggests possible solutions. GRAM (Grid Resource Allocation Manager) of Globus Toolkit 3.2 and relative technology on it is studied, and the solutions to extend GRAM to support task rescheduling and the cooperation type task are proposed. The design and implement of the GRAM client based on Web that is the kernel part of the task anagement module are discussed, and the design of resource discovery module and task transfer module are gived. Then, based on analysis of the Min-min scheduling algorithm, using QoS guided Min-min scheduling algorithm and the idear of categorizing computer resource as reference, the idear and implement of Balance-QoS guided Min-min algorithm model that does better in load balance and makespan are proposed. Finally, the design and implement of scheduling algorithm verification module are discussed, and the contrast of scheduling performance between Balance-QoS guided Min-min scheduling algorithm and Min-min scheduling algorithm on the module is gived.