云环境下能耗感知的虚拟机动态迁移策略研究  

Energy-aware Strategy Using Virtual Machine Live Migration in Cloud Environment

作　　者：钱育蓉[1] 于炯[1] 卞琛[1] 英昌甜[1] 孙华[1] 

机构地区：[1]新疆大学软件学院,新疆乌鲁木齐830008

出　　处：《微电子学与计算机》2015年第9期138-143,共6页Microelectronics & Computer

基　　金：国家自然科学基金项目(61363083;61262088);新疆青年博士科技人才创新项目(2013731004);新疆高校科研计划项目(XJEDU2012I10);新疆自治区自然科学基金(2013211A011;2011211A011)

摘　　要：针对云计算服务环境下静态阈值的虚拟机(Virtual Machine,VM)动态迁移算法无法根据云任务负载实时优化系统能耗和负载均衡的问题,(方法)提出利用工作负载历史数据的统计学规律自适应地确定迁移阈值,实现云计算环境下虚拟机(Virtual Machine,VM)动态调度的自适应节能算法.(结果)采用动态电压频率调节(dynamic voltage frequency scaling,DVFS)--实现系统部件静态节能,又通过两种自适应主机超载判定(Adaptive Host Overloading Detection,AHOD)算法实现云数据中心的VM动态迁移,在Cloudsim云仿真平台下对比实现DVFS静态节能和2种AHOD动态节能策略,(结论)结果表明:DVFS和AHOD策略可以显著节能68%以上,虽然AHOD策略中VM动态迁移引起的服务等级协定违例(Service Level Agreement Violation,SLAV)将导致云服务质量(Quality of Service,QoS)降低,并造成数据中心综合性能下降,然而合理建模并找到能耗与QoS的平衡点,即安全阈值为1.2的AHOD算法不但能够实现自适应负载均衡的虚拟机动态迁移,还实现云环境下的高效节能.In cloud computing environment, Virtual Machine （VM） dynamic migration algorithm with static threshold could not optimize the energy consumption and load balancing with real-time cloud workload.In response to these problems,the proposed work defined the migration threshold adaptively using the statistical regularity of workload historical data, in order to achieve the adaptive energy-efficient VM scheduling algorithm in cloud computing environment.Dynamic voltage frequency scaling （DVFS） was adopted as the static energy saving strategy,while VM live migrations with 2 AHOD （Adaptive Host Overloading Detection）algorithms were adopted as the dynamic energy saving strategy in cloud data center.After the simulation and contrast to the static energy saving of DVFS and the dynamic energy saving with AHOD algorithms,the results show that：DVFS and AHOD strategy can significantly save energy 68% or more,although VM dynamic migration of AHOD strategy caused SLAV （Service the Level Agreement Violation）,which lead to the degradation of QoS （Quality of Service）and the overall performance of the data center in cloud computing.However reasonable modeling is able to balance the energy consumption and QoS,security threshold of 1.2 in AHOD algorithm could consume the energy efficiently, achieve the adaptive load balancing of virtual machines migration dynamically.

关 键 词：动态电压频率调节 服务等级协定违例 自适应主机超载判定 Cloudsim 仿真 LOESS 回归 最短迁移时间 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]