网格计算中任务调度算法的研究和改进

摘要

Task scheduling is the center of grid computing research. The aim of task scheduling is distributing tasks to achieve the optimal scheduling scheme and complete computing tasks effectively. In a grid environment how to use the resources effectively is one of the most important problems in task scheduling. The paper firstly introduces the design theory, advantages and disadvantages of the static task scheduling and dynamic task scheduling. The paper presents a new scheduling algorithm called SA-MM considering the advantages of the most classical Min-Min and Max-Min algorithms in grid computing. The SA-MM schedules the corresponding algorithm to map the tasks and resources according to the use of the resources. Finally, the Min-Min, Max-Min and SA-MM algorithms are simulated with the aid of the GridSim simulation toolkit. The paper analyzes and compares the performances which affect the efficiency of task scheduling including MakeSpan and the resource load of the three task scheduling algorithms.%任务调度一直是网格计算中的热点问题,任务调度的目的是最优地分配任务,实现最佳的调度策略,以高效地完成计算任务.在网格环境中,资源的合理有效利用是实现任务调度的关键问题之一.本文首先论述静态任务调度算法和动态任务算法的原理和优缺点等,然后结合Min-min、Max-min算法的优点设计一种新的调度算法SA-MM,根据资源的使用情况自适应调度相应算法进行任务到资源的映射.最后,用GridSim模拟工具对网格计算中Min-min、Max-min和SA-MM任务调度算法进行仿真实验,分析和比较它们的调度长度(MakeSpan)和资源负载情况等影响任务调度效率的指标.