计算机应用与软件Issue(9)：208-210,214,4.DOI:10.3969/j.issn.1000-386x.2013.09.057
Hadoop平台下计算模型中调度策略的研究
RESEARCH ON SCHEDULING STRATEGY OF COMPUTING MODEL BASED ON HADOOP PLATFORM
曹旭 1张云华1
作者信息
- 1. 浙江理工大学信息学院 浙江 杭州310018
- 折叠
摘要
Abstract
On Hadoop platform , the default scheduling algorithms , FIFO, computing capacity scheduling and fair scheduling , all comply with strict queue order in scheduling process , this results in the problem that some of the tasks are scheduled to the nodes not satisfying the local property of data .In light of the problem , we propose a local property-based scheduling algorithm called delay scheduling .While maintaining the principle of fairness , the algorithm will arrange a job to be scheduled but cannot launch a local task to wait for a small amount of time but let other jobs launch the tasks instead .Experimental result demonstrates that , this algorithm shortens the average response time of the jobs, effectively increases the throughput of the cluster system , and raises the utilisation rate of the cluster resource .关键词
Hadoop/作业调度/延时调度/本地性Key words
Hadoop/Job scheduling/Delay scheduling/Local property分类
信息技术与安全科学引用本文复制引用
曹旭,张云华..Hadoop平台下计算模型中调度策略的研究[J].计算机应用与软件,2013,(9):208-210,214,4.
