文献类型：期刊文献

中文题名：基于马尔可夫决策过程的云平台资源调度

英文题名：Markov Decision Processes Based Resource Scheduling in Cloud Environment

作者：邱远[1];虞慧群[1];范贵生[1]

机构：[1]华东理工大学计算机科学与工程系,上海200237

年份：2016

卷号：42

期号：5

起止页码：702

中文期刊名：华东理工大学学报（自然科学版）

外文期刊名：Journal of East China University of Science and Technology

收录：CSTPCD;;Scopus;北大核心:【北大核心2014】；CSCD:【CSCD2015_2016】；

语种：中文

中文关键词：云计算;资源调度;马尔可夫决策过程;鲁棒性

外文关键词：cloud computing; resource scheduling; Markov decision processes; robustness

摘要：云计算平台可以动态地配置资源,适合基于工作流的科学计算。当前云平台的资源调度研究更多考虑运行时长和成本的最优化,而较少提到鲁棒性。本文提出了一种基于马尔可夫决策过程理论的资源调度算法,对工作流任务进行分组,按照任务的计算量和依赖关系将任务期限分配给各个任务组,在满足工作流总期限的基础上,将异构环境中的云资源分配给工作流的各个任务,通过最大化每个任务组的容忍时间使得整个工作流的鲁棒性达到最优。实验结果表明:该调度算法在异构环境中可以在任务期限和开销内提高调度的鲁棒性。
The characteristic of provisioning resource dynamically in cloud platform is suitable for the scientific computation of workflows. The existing works on workflow scheduling mainly consider the factors of makespan and cost optimization, and have little involving in robustness. In this paper, a resource scheduling algorithm based on Markov decision process theory is proposed,in which the tasks of the whole workflow are partitioned and the deadline on task partitions based on calculation time of task is distributed. Under the requirement that the deadline of whole workflow is not violated, the proposed algorithm makes the tolerance time maximum, and finally allocates resources on workflow tasks in heterogeneous cloud environment. The experimental results show that the proposed algorithm could effectively improve the robustness of the schedule within a given deadline and budget.