摘要
该文在Hadoop实现的MapReduce架构基础上,分析了现有的三种作业调度算法,针对当前算法没有考虑服务器负载状况和数据本地性差的缺点,提出了基于可变长度队列的公平调度算法(FSVQ),该算法分析了空闲节点率,并通过采取等待的办法满足考虑数据本地性。实验证明该算法可增加服务器集群的工作效率,减少网络延迟,具有实际的应用意义。
Based on the MapReduce framework achieved by Hadoop,this article has a detailed analysis of the existing three job scheduling algorithm, For these three algorithms does not consider the load conditions of server cluster and poor data locality, this paper presents the FSVQ algorithm which analyze the number of the arrival of idle nodes in certain time,it is also meet the data locality by waiting approach.The results prove that the algorithm can increase the efficiency of server clusters and reduce network latency, with practical application significance by test.
出处
《电脑知识与技术（过刊）》
2014年第5X期3206-3208,3211,共4页
Computer Knowledge and Technology
