基于分组遗传算法的数据中心虚拟机节能映射

近年来,随着人们对云计算业务需求持续增长,数据中心能耗日益增加,由此不仅增加了运营成本,巨大的 碳排放对生态环境也产生严重的影响,数据中心节能已成为当前亟须解决的重要难题。 对云数据中心的虚拟机放 置(Virtual Machine Placement, VMP)进行优化能有效地提高资源利用率,同时,VMP 也是减少数据中心能耗的重 要技术之一;针对数据中心的能耗感知 VMP 问题,提出一种基于分组遗传算法的节能算法 EEGGA (Energy - Efficient Grouping Genetic Algorithm),算法将节能 VMP 问题视为装箱问题(Bin Packing Problem,BPP),并应用基于 分组编码的遗传算法对其进行求解,通过减少活动物理主机的数量(装箱数量)以实现降低数据中心能耗的目标; 在算法迭代过程的交叉和变异等阶段,设计了多种启发优化策略提升子代染色体的适应度,从而提高算法的节能 性能和加快迭代收敛的速度;通过仿真实验,在收敛速度和求解性能等方面将提出的算法与传统的节能遗传算法 进行对比,实验结果表明:提出的算法能够有效地减少数据中心的能耗,在节能性能和求解收敛速度方面均优于其 他算法。

An Energy-efficient Virtual Machine Mapping in Data Centers Based on Grouping Genetic Algorithm

In recent years with the continuous growth of the demand for cloud computing the energy consumption of cloud data centers has been increasing which not only brings economic problems but also has a serious impact on the ecological environment caused by huge carbon emissions. Therefore data center energy saving has become an important problem to be solved urgently. In cloud data centers Virtual Machine Placement VMP optimization can effectively improve resource utilization. VMP is also one of the important technologies to reduce the energy consumption of data centers. Aiming at the problem of energy-aware VMP of cloud data centers an Energy-efficient Grouping Genetic Algorithm EEGGA based on grouping genetic algorithm GA was proposed. The algorithm considered the energy-efficient VMP problem as a bin packing problem BPP and applied a genetic algorithm based on grouping coding to solve it. The goal of reducing the overall energy consumption of data centers was achieved by reducing the number of active physical hosts the number of containers . At the same time several heuristic optimization strategies were designed to improve the fitness of offspring chromosomes in the crossover and mutation stages of the iteration process so as to improve the energy-saving performance of the algorithm and accelerate the speed of convergence. In the simulation experiments the proposed algorithm was compared with traditional algorithms in the aspects of convergence speed and energy-saving performance. Experimental results show that the proposed algorithm can effectively reduce energy consumption and is superior to other algorithms in terms of energy-saving performance and convergence speed.