基于改进粒子群算法的虚拟机放置策略研究  

Research on Virtual Machine Placement Strategy Based on Modified PSO Algorithm

作　　者：唐忠原 何利文[1] 黄俊[1] 袁野[1] TANG Zhong-yuan;HE Li-wen;HUANG Jun;YUAN Ye(School of Computer Science,Nanjing University of Posts and Telecommunications,Nanjing 210046,China)

机构地区：[1]南京邮电大学计算机学院,江苏南京210046

出　　处：《计算机技术与发展》2018年第7期93-98,共6页Computer Technology and Development

基　　金：江苏省"六大人才高峰"高层次人才项目(2014-WLW-005);南京邮电大学引进人才科研启动基金(NY212012);中兴通讯研究基金(2015外)

摘　　要：大规模虚拟机放置是云计算中的一个关键问题,好的虚拟机放置策略不仅可以提高数据中心的性能、资源利用率,同时也可以减少数据中心的能耗与维护费用。当前,虚拟机放置研究的重要方向是数据中心资源的多目标优化。鉴于此,提出以网络通信的优化与降低能耗为目标,以服务器的CPU、内存与带宽等资源为约束条件的系统化部署方案。在降低能耗方面,不仅考虑到了服务器的能耗,同时也考虑了通信设备的能耗。提出了一种改进粒子群算法的虚拟机放置策略,首先对原始粒子群算法的参数进行相应的重定义或修改使其适合解决离散优化问题,再通过突变策略增强粒子群的多样性以避免微粒因早熟而陷于局部最优的问题。实验结果证明,该放置策略可以使数据中心的能耗降低6%~20%,同时也可使通信的总流量与主干网流量降低5%~30%。Large-scale virtual machine placement in the cloud computing is an important problem that remains to be effectively addressed.The excellent virtual machine placement strategy can enhance performance and resource utilization,and reduce power consumption and maintenance cost for data centers.At present,the important direction of the research on virtual machine placement is the multi-objective optimization of data center resources.In viewof this,we propose an systematical scheme with the objective of the optimization of network communication and the reduction of energy consumption,and the server＇s CPU,memory and bandwidth and other resources as the constraints.In terms of lower energy consumption,we take both the server energy consumption and the energy consumption of the communication device into account.After that,we present a virtual machine deployment strategy based on modified particle swarm optimization（ MPSO）.The original PSO must be improved including redefining or modifying parameters to solve the problem of discrete optimization.Then,the diversity of the particle swarm is enhanced by the mutation strategy to avoid the problem that the particle is trapped in the local optimum due to the premature.Experiment indicates that this strategy can not only cuts the energy consumption by 6 ~ 20 percent,but also reduces the total traffic of communication and main network traffic by 5 ~ 30 percent.

关 键 词：云计算 虚拟机放置 粒子群算法 突变 能耗 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]