基于动态规划的云计算任务调度研究  

Cloud computing task scheduling research based on dynamic programming

作　　者：史少锋[1] 刘宴兵[1] 

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2012年第6期687-692,共6页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金："新一代宽带无线移动通信网"国家科技重大专项课题(2011ZX03002-004-03);教育部NCET;重庆市高等教育成果转化项目(Kjzh10206);公安部信息网络安全重点实验室项目(C11609)~~

摘　　要：任务调度作为云计算中的核心问题,其目的是合理分配任务,实现最佳调度策略并且有效地完成任务。首先对当前云计算中任务调度算法进行分析并指出传统算法的不足与缺点,然后提出一种基于动态规划模型的任务调度算法,以任务运行时间最少为优化目标,把任务与数据集群中虚拟机匹配看成为多阶段决策的组合优化。最后在CloudSim进行仿真实验,结果表明,所提算法在满足多用户需求下与max-min和min-min算法相比,在一定数量规模下,任务完成时间减少而资源负载相对均衡。Task scheduling is the core issue in cloud computing system,the aim of task scheduling is reasonably distributing tasks to achieve the optimal scheduling scheme and complete tasks efficiently.In this paper,we have firstly analyzed existing task scheduling algorithms in cloud computing and pointed out the shortages and disadvantages of the traditional algorithms,then under these conditions we propose a task scheduling algorithm based on dynamic programming model.We think the match between tasks and virtual machines in the datacenter cluster is a problem about combination optimization of multi-stage decision-making,and its objective is the least execution time.In the end,max-min,min-min algorithms and the proposed algorithm are simulated with the aid of the CloudSim toolkit.The result demonstrates that the proposed algorithm has better performance in terms of task completing time and resource load than the classical algorithms.

关 键 词：云计算 任务调度 动态规划 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]