摘要
为解决Hadoop云平台下作业无法满足时间约束的问题,提出一种基于资源预估的作业调度算法。通过建立资源预估模型计算作业所需资源,结合作业间的资源竞争关系对完成时间进行判定,最后根据作业的数据本地性改进延迟调度策略。实验结果表明,该算法能够满足作业对时间约束的需求,提升系统的资源利用率。
In order to satisfy the deadline constraints in Hadoop, this paper proposed a job scheduling algorithm based on data- aware. This algorithm proposed a resource estimation model and computed job＇ s deadline, then optimized the delay strategy ac- cording to resource locality. Through analyzing the result of experiments, this algorithm can satisfy the deadline constraints and improve the efficient in Haoop.
出处
《计算机应用研究》
CSCD
北大核心
2016年第8期2311-2314,共4页
Application Research of Computers
基金
四川省科技厅基金资助项目(2012FZ0081)
