自适应调整虚拟机权重参数的调度方法  

An Adaptive Scheduling Method of Weight Parameter Adjustment on Virtual Machines

作　　者：王凯[1,2,3] 侯紫峰[1,3] 

机构地区：[1]中国科学院计算技术研究所,北京100190 [2]中国科学院研究生院,北京100190 [3]联想研究院,北京100085

出　　处：《计算机研究与发展》2011年第11期2094-2102,共9页Journal of Computer Research and Development

摘　　要：在基于特权服务操作系统的虚拟机架构下客户操作系统需要借助特权服务操作系统来访问真实硬件,目前虚拟机调度算法的优化主要是侧重于I/O密集型虚拟机的研究,而忽视了CPU密集型虚拟机,更忽视了特权服务操作系统的I/O处理能力对虚拟机整体性能的影响.针对这些问题,提出了一种基于Credit算法的自适应调整虚拟机权重参数的优化调度方法,将特权服务操作系统的I/O处理能力作为虚拟机参数调整的一个重要参数,同时兼顾I/O密集型虚拟机和CPU密集型虚拟机对资源的需求.实验结果表明该方法能够及时根据当前的I/O请求数量和特权服务操作系统的处理能力合理调整虚拟机的权重参数,从而大大提高了客户操作系统CPU处理性能和硬件设备的访问性能.In the virtual machine architecture including Service OS and Guest OSs, Guest OS can access the real hardware by the aid of Service OS. But the optimizations of the current scheduling algorithms are focused on I/O-intensive domains. They neglect CPU-intensive domains and the effect of the I/O processing capacity of Service OS on the whole performance of virtual machines. Aiming at these problems, an adaptive scheduling method of parameters on virtual machine based on credit scheduling algorithm is given in this paper. It takes the I/O processing capacity of Service OS as an important parameter and pays attention to I/O-intensive domain and CPU-intensive domain, which is advantageous to make system resources get used more reasonably. The experiment results show that the proposed method can make the performance of CPU processing and accessing real hardware of Guest OS better than that of some given ones, such as Credit scheduling algorithm, through the reasonable adjustment of virtual machines＇ weight according to the current I/O requests of Guest OSs and the processing capacity of Service OS. The scheduling algorithm of virtual machines has direct effect on the performance, and it is a significant job to go into scheduling algorithms of virtual machines.

关 键 词：虚拟机监控器 特权服务操作系统 客户操作系统 权重 处理能力 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]