异构云环境下AHP定权的多目标强化学习作业调度方法
作者:
作者单位:

1.武汉理工大学计算机科学与技术学院;2.上海交通大学 电子信息与电气工程学院

作者简介:

通讯作者:

中图分类号:

TP311

基金项目:

国家自然科学基金项目(面上项目,重点项目,重大项目);湖北省创新团队项目;湖北省技术创新专项重大项目


Multi-Objective Reinforcement Learning Job Scheduling Method using AHP Fixed Weight in Heterogeneous Cloud Environment
Author:
Affiliation:

1.Wuhan University of Technology,School of Computer Science and Technology;2.Shanghai Jiao Tong University,School of Electronic Information and Electrical Engineering

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    随着新型基础设施建设(新基建)的加速,云计算将获得新的发展契机.数据中心,作为云计算的基础设施,其内部服务器不断升级换代,这造成计算资源的异构化.如何在异构云环境下,对作业进行高效调度是当前的研究热点之一.本文针对异构云环境多目标优化调度问题,设计了一种AHP定权的多目标强化学习作业调度方法.首先我们定义了执行时间、平台运行能耗、成本等多个目标.其中定义服务延迟成本用来描述用户对服务质量的满意程度.然后设计了面向异构资源的多目标调度综合评价方法,该方法利用层次分析法(Analytic Hierarchy Process, AHP)确定了各个目标的权重.最后将该方法引入Q-learning的奖励值计算,使其能反映异构云环境下作业的总体执行情况,并对后续抵达的作业起到良好的经验学习作用.实验结果表明本文提出的方法优于大部分对比方法,能较好地优化作业执行效率和保障用户及服务提供商的利益.

    Abstract:

    With the acceleration of new infrastructure, cloud computing will be given an entirely new opportunity to develop. Data center, as the infrastructure of cloud computing, its internal servers are continuously updated, which leads to the heterogeneity of computing resources. How to efficiently schedule jobs in heterogeneous cloud environment has become an increasingly popular research. In this paper, we design a multi-objective reinforcement learning job scheduling method using AHP fixed weight in heterogeneous cloud environment. First, we define the execution time, energy consumption, execution cost, etc. Service delay cost is used to describe the user satisfaction for service. Then, a comprehensive evaluation method for multi-objective scheduling is designed. The weight coefficient of each object is determined by the analytic hierarchy process (AHP). The method is introduced into the calculation of reward, makes it reflective of the overall situation and serves as an excellent learning tool for the following jobs. The experimental results show that the proposed method can better optimize the execution efficiency, while ensuring the interests of users and service providers.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2020-07-06
  • 最后修改日期:2021-07-28
  • 录用日期:2020-12-03
  • 在线发布日期: 2021-01-04
  • 出版日期: