Job Scheduling and Processor Allocation in Computing Grid

Abstract

Grid computing aims at integrating computing resources located at different places to share computing power and workload as well as to solve a large computational problem collaboratively. It can achieve load sharing and improve resource utilization. Workload management is a key factor to achieve the goal. Workload management can be viewed as two collaborative parts: job scheduling and resource allocation. We developed a simulation environment to study the resource fragmentation issue on grid platforms. We tried to compare and analyze different scheduling and allocation methods for finding out the method which can reduce resource fragmentation occurrences most. We developed a new resource allocation method called Most-Fit, which was shown superior to other methods under several workload conditions. The simulation studies in this thesis indicate SJF (Smallest Job First) scheduling method together with the Most-Fit allocation method outperformed other combinations of methods under most workload conditions.