基于粒子群优化算法的集群调度策略

Scheduling Strategy Based on Particle Swarm Optimization Algorithm

针对集群调度问题的特点,设计了基于粒子群优化算法的调度策略.与传统backfill算法相比,粒子群优化算法对作业比较公平,能避免对大作业响应慢的缺点,使得调度策略在生成速度和精度上都有明显的提高.实验结果表明,该调度策略能较好地提高CPU利用率和缩短作业平均响应时间.

In cognizance of the characteristics of cluster scheduling problem, scheduling strategy based on particle swarm optimization is designed and implemented. Compared with backfill algorithm, PSO algorithm can better improve the fairness of jobs. It can avoid the problem that bigger jobs can＇t be executed quickly. The speed and accuracy of strategy generation are improved significantly. The experiment results show that the algorithm increases the utilization of the CPU and reduces average response time.