虚拟机聚合对数据中心温度及功耗影响的研究

Research on effect of VM consolidation to Internet data centers' temperature and power

负载聚合特别是虚拟机聚合是降低数据中心计算功耗、减少数据中心运行成本的有效方法,传统虚拟机聚合方法在降低计算功耗时却忽略了对冷却功耗的考虑。根据数据中心温度功耗模型探讨了负载聚合的必要条件,提出了采用遗传算法来指导虚拟机聚合,优化系统整体功耗。以实际网上书店的访问数据进行因特网数据中心虚拟机聚合仿真实验,结果表明基于遗传算法的虚拟机聚合方法可以有效降低系统温度,并以微弱的性能下降代价大幅地降低了系统计算功耗和冷却功耗。

Workload consolidation especially virtual machine consolidation can greatly reduce working power and operating cost. Traditional VM consolidation methods try to minimize the computing power, while with little consideration about cooling cost. This paper first explored the＂ necessary condition for workload consolidation. Then it proposed a genetic algorithm to optimize the total energy cost. Finally it carried out VM consolidation simulations in an Internet data center environment with some online bookstore working traces. The experiment results show that the VM consolidation method with genetic algorithm can efficiently reduce the system peak inlet temperature and energy cost while with little performance reduction.