RESEARCH OF TASK SCHEDULING BASED ON PERFORMANCE UNDER GRID ENVIRONMENT

摘要

Grid computing is a special kind of branch network computing which has important innovative ideas and great potential of development. It forms a large use of each other and co-operation high-performance computing network by interconnecting a number of computing resources, data resources of the wide area network as well as other equipments. Since grid system is heterogeneous and dynamic, and the different resource requires of the applications running on grid system, grid task scheduling is extremely complicated. Therefore, how to syncopate the task will directly affect the overall scheduling performance of the task in a grid system. In implementation process, different tasks need different resource, this article mainly aims at the demand of CPU and memory to establish a performance evaluation model. Through the evaluation results of CPU and memory, establish a rational task segmentation strategy to improve the quality of service, and realize grid load balancing, low communication complexity and high reliability between tasks, which achieve the purpose of minimizing the task execution time.