基于PSO算法的网格任务调度策略  

Task scheduling in grid based on PSO algorithm

作　　者：迟玉红[1,2] 白鹏[2] 于俊发[2] 喻春明[2] 

机构地区：[1]清华大学计算机科学与技术系 [2]中国人民解放军65053部队

出　　处：《辽宁工程技术大学学报（自然科学版）》2010年第2期274-277,共4页Journal of Liaoning Technical University (Natural Science)

基　　金：国家重点基础研究专项基金资助项目(G2007cb311003);国家自然科学杰出青年基金资助项目(60625304)

摘　　要：为了合理地协调和分配网格资源,降低网格任务完成时间,有效保持网格资源负载平衡,通过分析网格任务调度的关键问题和PSO算法的优化原理,建立了网格任务调度的元任务模型和性能指标的数学模型,确定了粒子编码和解码方式,提出了一种基于局部模型PSO算法的网格任务调度策略,并在PSO算法处理粒子边界问题时,采用了"圆桌运动"的新方法。仿真实验结果表明,算法可行有效,并且改善了PSO算法易陷入局部最优的问题。It is a challenge to find the optimal solution which can rationally coordinate and allocate the resources in grid system,greatly reduce the completion time,efficiently balance workload and improve grid performance.This paper investigates the key issues of task scheduling and the principle of PSO algorithm,develops the model of meta task in grid system and the model of performance metrics of task scheduling.Also the paper has developed a new method for coding and encoding of particles,and proposed a task scheduling scheme based on PSO algorithm using local best model.To give the boundary conditions,the paper proposes a new method called ＂round-table moving＂.The experimental results show that the proposed approach contributes to overall grid load balancing,significantly improves the grid application execution performance and resource utilization,and avoids the PSO＇s ＂premature＂ problem.

关 键 词：网格 任务调度 元任务 PSO算法 局部模型 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP393[自动化与计算机技术—控制科学与工程]