Computer Engineering and Applications ›› 2014, Vol. 50 ›› Issue (4): 86-89.

Previous Articles     Next Articles

Improved LATE scheduling algorithm on Hadoop paltform

HU Dan, YU Jiong, YING Changtian, ZOU Weiming   

  1. College of Information Science and Technology, Xinjiang University, Urumqi 830046, China
  • Online:2014-02-15 Published:2014-02-14

Hadoop平台下改进的LATE调度算法

胡  丹,于  炯,英昌甜,邹伟明   

  1. 新疆大学 信息科学与工程学院,乌鲁木齐 830046

Abstract: The existing schedulers of MapReduce on the Hadoop platform are analyzed. To consider the lack of LATE scheduling algorithm in allocating TaskTracker to execute backup tasks for backward tasks, combining with the particularity of Hadoop cluster heterogeneity and workload, an improved scheduling algorithm is proposed based on LATE scheduling algorithm. The experiments and performance analysis for the algorithm show that it has significantly improved in the completion time and load balancing.

Key words: Hadoop, MapReduce, LATE, backward task

摘要: 对Hadoop平台下的MapReduce现有的调度器进行分析研究。针对LATE调度算法在分配节点执行落后任务的备份任务时的不足,结合Hadoop集群的异构性和工作负载的特殊性,在LATE调度算法的基础上提出了一种改进的LATE调度算法。对该算法进行实验和性能分析,表明该算法在完成时间和负载均衡方面有很大改进。

关键词: Hadoop, MapReduce, LATE, 落后任务