云数据中心高能效的虚拟机放置算法

An Energy-efficient Virtual Machine Placement Algorithm in Cloud Data Center

面对云数据中心高能耗的挑战,以节能为目标的虚拟机放置算法成为研究热点.现有研究大多只考虑CPU一种系统资源对能效的影响,并且多采用基于贪心算法的传统启发式算法进行虚拟机放置.已有研究开始考虑多种系统资源对能效的影响,并且提出了多资源的能效模型,但是在多资源能效模型下虚拟机放置算法的研究还未引起关注.本文根据多资源的能效模型提出了基于粒子群算法的高能效虚拟机放置算法,包括采用首次适应算法生成粒子,定义粒子个体最优解和全局最优解的信任度指导粒子进化,根据多资源能效模型定义粒子群的适应度函数并以此来评价粒子.仿真实验结果表明,与传统启发式算法相比较,该算法使虚拟机的放置结果更接近系统能效的最佳状态,同时也有效地提高了系统资源的利用率.

Facing the challenge of high energy consumption in cloud data center, virtual machine placement algorithms aiming at ener- gy efficiency become a research hotspot. Previous investigations mostly have focused on the effects of CPU on energy efficiency. In addition these studies exclusively use traditional heuristic algorithms based on greedy method to place virtual machines. Some investi- gations have taken multiple system resource related factors into consideration for energy efficiency, and a multi-resource efficiency model has also been proposed. However, the study of virtual machine placement algorithm based on the multi-resource efficiency mod- el has not caused concern. In this paper, we propose an energy-efficient virtual machine placement algorithm for multi-resource energy efficiency. The method is developed based on particle swarm optimization algorithm. It uses First Fit algorithm to generate particles, defines the trust for the individual optimal solution and the global optimal solution of particles to guide particles＇ evolution, defines fit- ness function of Particle Swarm Optimizer,and evaluates particles according to the energy efficiency model. Simulation experimental results find that, compared to traditional heuristic algorithms, the current algorithm results in a virtual machine placement that approa- ches the best for energy efficiency. Moreover, it improves significantly the utilization of system resources.