基于马尔科夫博弈的云代理与微云收益优化  

Optimal revenue of Cloud Broker and Cloudlet based on Markov games

作　　者：张锋辉 符茂胜 何富贵 ZHANG Feng-hui;FU Mao-sheng;HE Fu-gui(School of Electronics and Information Engineering, West Anhui University, Liu ’an 237012, China)

机构地区：[1]皖西学院电子与信息工程学院,安徽六安237012

出　　处：《计算机工程与设计》2018年第12期3628-3632,共5页Computer Engineering and Design

基　　金：国家自然青年科学基金项目(61702375);安徽省高校优秀青年人才基金项目(gxyqZD2016249;gxyqZD2016199);安徽省高校自然科学研究重大基金项目(KJ2015ZD44);皖西学院校级基金项目(2013LW003;KJ103762015B09)

摘　　要：在移动云计算中提高微云(Cloudlet)虚拟机利用率是增加云代理(Cloud Broker)和微云收益的关键,引入马尔科夫博弈可分别优化双方的收益。将云代理和微云和的服务量化为排队过程,分析该过程建立马尔科夫博弈模型;分析每个时间片中系统的收益并确定该博弈为变和马尔科夫博弈,提出反向迭代算法得到其纳什均衡策略。对比马尔科夫博弈的方法与云代理租用确定数量虚拟机的方法,其结果表明,采用马尔科夫博弈可明显提高系统收益。Improving the utilization of virtual machines in Cloudlet is key to achieve higher benefits of Cloud Broker and Cloudlet,using Markov game can optimize the benefits of both sides.The services of Cloud Broker and Cloudlet were quantized as queueing process.Markov game was constructed through analyzing this model.The changeable sum Markov game was defined through analyzing the revenue of this system,the Nash equilibrium was reached using backward iterative algorithm.The Markov game model was compared with the approach that Cloud Broker Rent determined virtual machines.The results show that the Markov game can significantly improve system revenues.

关 键 词：马尔科夫博弈 移动云计算 云代理 微云 纳什均衡 收益优化 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]