一种基于K均值聚簇的虚拟机分类与部署方法

A VIRTUAL MACHINE CLASSIFICATION AND PLACEMENT METHOD BASED ON K-MEANS CLUSTERING

云数据中心环境下,虚拟机部署结果对主机能耗与服务等级协议SLA的遵守均具有重要影响。为了降低数据中心能耗与SLA违例,提出一种基于三门限值的高能效虚拟机部署优化算法。基于历史数据集,设计一种中档四分位的K-均值聚簇方法以产生主机CPU利用率的三个门限值;依据三个门限值,将主机划分为低载主机、轻量负载主机、正常负载主机和重载主机四种类型;为了对重载主机实施虚拟机迁移,分别针对计算密集型任务和I/O密集型任务设计两种虚拟机迁移选择方法,实现虚拟机优化部署;通过现实负载流数据对算法进行仿真分析。结果表明,该算法不仅可以有效降低能耗,而且SLA违例也较低,相比单纯降低能耗而忽略性能的同类算法,具有更高的能效。

In cloud data center environment,virtual machine placement greatly affects on the energy consumption of hosts and complied with Service Level Agreement(SLA).To reduce the energy consumption and SLA violation,we proposed a high energy-efficient virtual machine placement optimization algorithm based on three thresholds.Based on the historical data set,we designed a K-means clustering algorithm with Midrange-Interquartile range to produce three thresholds of CPU utilization on hosts.According to the three thresholds,we divided all hosts into four types: hosts with little load,hosts with lightly load,hosts with moderate load and hosts with heavy load.Then,for migrating some virtual machines from the hosts with heavy load,we designed two kinds of virtual machine migration selection algorithms for the computation intensive tasks and the I/O intensive tasks respectively.Finally,we performed some simulation experiments using the real-world workload.The results show that our algorithm not only can reduce the energy consumption,but can maintain low SLA violation.Compared with the same type of algorithms that only reduce the energy consumption without considering the performance,our algorithm can get higher energy-efficiency.