基于思维进化的集群作业调度方法研究  

Research on Mind Evolutionary Computation Based Job Scheduling for Server Clusters

作　　者：薛正华[1] 刘伟哲[1] 董小社[1] 伍卫国[1] 

机构地区：[1]西安交通大学计算机科学与技术系,西安710049

出　　处：《西安交通大学学报》2008年第6期651-654,764,共5页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(60773118);国家高技术研究发展计划资助项目(2004AA111110;2006AA01A109)

摘　　要：为了减少不准确的时间估计对作业调度的影响,提出了动态预约调度模式.在该模式中,预约队列中的作业与被预约资源是松耦合的,可根据作业完成情况进行重映射,以减少作业提前完成而产生的资源碎片.基于动态预约模式,将思维进化计算引入到作业调度中,以各种回填算法的调度结果作为初始群体,通过趋同和异化操作,使群体不断向最优解进化,从而产生更优的调度方案.基于真实作业集的仿真结果表明,所提算法的作业平均的等待、延迟时间比最优的基于回填的组合算法分别下降了68.5%和66.9%.A dynamic reservation based on scheduling method is proposed to reduce the disadvantage caused by inaccurate estimation on job run time. The method combines the reservation jobs and the computing resources in loose couple, and it enables the scheduler to recombine the reservation jobs and resources to reduce the idle resources caused by jobs completed in advance. Based on the dynamic reservation mode, the mind evolutionary computation （MEC） based algorithm is introduced into the scheduler. The algorithm sets the result obtained from the backfill algorithms as the initial population, and it can achieve better scheduling result by adopting assimilation and dissimilation operations. Simulations based on a real workload show that the proposed method is valid. Compared with the backfill algorithm, the proposed algorithm decreases both the average job wait time and the average slowdown by 68. 5% and 66.9%, respectively.

关 键 词：服务器集群 作业调度 思维进化计算 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]