基于遗传算法的自适应网格任务调度方法

An Adaptive Grid Job Scheduling Method Based on Genetic Algorithm

文章提出了一种以资源代理为基础的任务调度方法—GMBSA,该方法先对任务执行时间进行预测,然后运用遗传算法结合多队列Backfilling方法进行任务调度,达到最小化任务执行时间(MinimumExecutionTime)的要求,最终实现网格资源的优化分配。试验中采用Simgrid任务调度模拟器对GMBSA的性能进行了测试,并比较了轻重负载情况下GMBSA,多队列Backfilling和FCFS三种调度方案的性能差异。

This paper presents a job scheduling method based on resource broker--GMBSA.Firstly,the method predicts the run times of job,then uses genetic algorithm combining with multiple-queue Backfilling to realize job scheduling,and optimizes the grid resources allocation.It uses Simgrid simulator to evaluate the performance of GMBSA in the experiments,and compares its performance with multiple-queue Backfilling scheduling algorithm and FCFS scheduling policy.