The Research And Implementation Of Task Scheduling Algorithm In Grid Computing

Grid computing technology has emerged as an important new field in international computer field. The goal of grid is to share resources, support remote and parallel access to computing resources, and to link geographically distributed kinds of resources with Internet to a single system image with high performance computing and information service environment. Because of the characters of distributed, dynamic, diverse, autonomous and manifold management of a grid, how to schedule tasks in grid environment to meet users' requirements is a challenging problem.Task scheduling technology is a part of grid core services technology. Task scheduling in grid computing resolves how to map reasonably grid resources to tasks and schedule it to those resources to execute. An advanced scheduling policy in favor of reducing total run time and total expense, by this means can enhance entire grid system's capability.The main research content of this paper includes three parts:(1) Three popular grid infrastructures and the difference of them are researched; then typical grid resource management model and systems are introduced.(2) Some classical task scheduling algorithms are discussed, focused on Min-Min algorithm. The algorithm is one of the most basic heuristic algorithms in the grid task scheduling. It is based on assigning each task to earliest available and fastest resource as soon as possible. It's defect is to lead to unbalanced load, lower use rate of resources and longer execution time. Aimed at the defect, the improved Min-Min algorithm is presented based on priority of tasks. The basic thought of the algorithm is that unassigned tasks first are descended arrangement according to the tasks' length, and divided into certain sections. Longer tasks have higher priority, first scheduled to best resources and shorter ones have lower, assigned to second resources, so the total time of executing assignment will be shortened.(3) Because building an actual grid test bed is both time-consuming and expensive, moreover needs various resources, using actual grid system that confirms the validity and the performance of scheduling algorithm is quite difficult. usually emulators are used to handle with the work. In the paper, several grid simulators are discussed. GridSim is introduced in details including it's architecture and three scheduling algorithm based on economic model. Finally, with the help of GridSim, the Min-Min algorithm and mended Min-Min scheduling algorithm are designed with java language, and performances of two algorithms compared.To the schedule of a large number of tasks, the results of this experiment show that Min-Min arithmetic based on tasks' priority is feasible and has lower total execution time than the original algorithm.