Abstract: Research of evaluation method upon virtual machines performance isolation degree is of great meaning to meet customers’ SLA and schedule virtual machines in virtualization data centers. To comprehensively evaluate virtual machines performance isolation degree, based on measuring, this paper proposes virtual machines performance isolation degree evaluation method based on multi-type workload macroeconomic performance indicators and recorded relation data of performance isolation degree, monitorable resources utilization and number of VMs, upon which it performs regression analysis. Eventually isolation degree sensed scheduling algorithm is proposed and proved effective in experiment.

Key words: virtual machines consolidation, regression analysis, performance isolation-aware dispatching algorithm, performance isolation

摘要： 针对虚拟机用户服务水平保障问题，提出了一种基于性能隔离度的虚拟机调度方法。在实验测量的基础之上，定义了基于多类型负载宏观性能指标的虚拟机性能隔离度量化计算方法，并在实验中记录了性能隔离度与可监测的资源利用率及虚拟机数量的关系数据，关系数据分布显示了它们之间存在的线性趋势，因此通过回归分析及实验验证了它们之间的线性关系。基于该线性模型设计了隔离度感知调度算法，对虚拟机进行调度从而实现虚拟化系统性能的整体优化，仿真实验证明了该调度机制对多虚拟机调度的有效性。

关键词: 虚拟机整合, 回归分析, 隔离度感知调度算法, 性能隔离