Effcient Scheduling Heuristics for Independent Tasks in Computational Grids

Abstract

Grid computing is an extension to parallel and distributed computing. It is an emerging environment to solve large scale complex problems. It enables the sharing, coordinating and aggregation of computational machines to full the user demands. Computational grid is an innovative technology for succeeding generations. It is a collection of machines which is geographically distributed under different organisations. It makes a heterogeneous high performance computing environment. Task scheduling and machine management are the essential component in computational grid. Now a day, fault tolerance is also playing a major role in computational grid. The main goal of task scheduling is to minimize the makespan and maximize the machine utilisation. It is also emphasis on detection and diagnosis of fault. In computational grid, machines may join or leave at any point of time. It may happen that machine is compromised by an advisory or it may be faulty due to some unavoidable reason like power failure, system failure, network failure etc. In this thesis, we address the problem of machine failure and task failure in computational grid. Also, we have focused on improving the performance measures in terms of makespan and machine utilisation. A simulation of the proposed heuristics using MATLAB is presented. A comparison of our proposed heuristics with other existing heuristics is conducted. We also demonstrate that number of task completion increases even if some of the machine work efficiently in computational grid.