基于动态负载均衡的分布式任务调度算法研究  

Research on a distributed task scheduling algorithm based on dynamic load balancing

作　　者：朱虹宇[1] 李挺[2] 闫健恩[1] 张兆心[1] 

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001 [2]国家计算机网络应急技术处理协调中心,北京100029

出　　处：《高技术通讯》2014年第12期1261-1269,共9页Chinese High Technology Letters

基　　金：国家科技支撑计划(2012BAH45B01);国家自然科学基金(61100189;61370215;61370211);国家信息安全242计划(2014A085);山东省中青年科学家奖励基金(BS2011DX001)资助项目

摘　　要：为提高分布式平台的性能,研究了其任务调度方法。针对分布式任务调度容易引起节点间负载不均衡,从而造成系统资源浪费的问题,提出了一种基于动态负载均衡的分布式任务调度算法。该算法根据各节点的实时性能指标(包括CPU利用率、内存使用率、平均负载指标、网络状况)进行任务动态调度,使各节点的负载相对均衡。在分布式平台下进行了拓扑探测、DNS分布与配置探测实验并对不同算法的性能做了对比。实验结果表明,采用基于动态负载均衡的任务调度算法任务完成时间比轮询调度算法平均减少30%,比Min-Min调度算法平均减少17%。To improve the performance of distributed systems, the distributed task scheduling was studied. Considering that distributed task scheduling easily causes load imbalance between nodes, so causing the wasting of system resources, a distributed task scheduling algorithm based on dynamic load balancing was proposed. The algorithm conducts the dynamic task scheduling according to the real-time performance indicators of CPU utilization, memory utilization, average load and network performance of each node to make the load of each node relatively balanced. The experiments on topology detection, DNS distribution detection and DNS configuration detection were performed under a distributed system to compare the performance of the proposed algorithm with other two algorithms. The results showed that the task completion time of the proposed algorithm was reduced by 30% compared to the polling sched- uling algorithm, and was reduced by 17% compared to the Min-Min scheduling algorithm.

关 键 词：分布式平台 任务调度 负载均衡 任务重组 

分 类 号：TP393.01[自动化与计算机技术—计算机应用技术]