The Grid Task Scheduling Optimization Method Research Based On Hypergraph

With modern engineering design, manufacturing production, life application are gradually becoming informatization, the demand of large-scale parallel computing is increasing. Grid computing,as a new technology industries, connecting computer, instruments of equipment, database and etc through the Internet, provides a application platform with sharing resources comprehensive for people, which meets the demand of the growing computing power of people. Grid computing is equivalent to the world’s largest computer, this paper study on the grid task scheduling which is in view of the massive computing tasks, researching assign task to computing resources reasonably.We study the grid task scheduling from two characteristics of it, which are huge resources and fault-tolerance of scheduling. Studying for one of the characteristics of grid scheduling which is huge resources. In order to improve the efficiency of searching resources on scheduling, clustering of pretreatment on resources with genetic algorithm. Clustering resources is based on the resource model which is built with hypergraph theory, transformed multi-objective into single objective with the characteristics of resources. With the minimum execution time as the main target, and loading balancing resources, a task scheduling algorithm based on multiple-objective optimal resource clustering(MORC) algorithm is designed; Studying for another one, fault-tolerance of scheduling, considering the complex task, primary-backup is used basing on the task model which is built with hypergraph theory. According the importance of the task and resource security to set up the dynamic replication, with the minimum execution time of task, a dynamic replication of fault-tolerant scheduling(DRFT) algorithm is designed. Finally, the simulation results are analyzed and compared of the two algorithms and classical algorithms. And applying these two scheduling algorithm in the aerodynamic noise calculation of aircraft landing gear. The results shows that, under the background of two experiments, the two algorithm in this paper are superior to the classical scheduling algorithms obviously in terms of computational efficiency, and other performance indicators.In this paper, based on the research of grid task scheduling, the study of hypergraph theory, a multi-objective genetic algorithm in clustering, dynamic replication fusing in task scheduling algorithm, especially building task and resource models with hypergraph theory, these play a supplementary role for MORC algorithm and DRFT algorithm design. In this paper, two kinds of algorithm can not only reduce the scheduling and execution time, also can improve the load balancing and fault tolerance performance of grid task scheduling.