一种灵活高效的虚拟CPU调度算法

发布时间：2018-01-10 11:21

本文关键词：一种灵活高效的虚拟CPU调度算法　出处：《软件学报》2017年02期 　论文类型：期刊论文

【摘要】：目前,虚拟化已经广泛应用于数据中心,但主流的虚拟CPU调度策略并没有实现对I/O性能的保障,尤其是当延时敏感型负载的虚拟机和计算敏感型负载的虚拟机竞争CPU资源时,其性能显著下降.针对上述问题,提出了一种灵活、高效的虚拟CPU调度算法FLMS(flexible I/O latency and multi-processor sensitive scheduler).FLMS通过采用虚拟机分类、虚拟CPU绑定、多类时间片等技术降低了虚拟机的响应延时,同时基于多处理器架构重新设计了负载均衡策略,优化了虚拟CPU迁移.FLMS通用于目前主流的虚拟化方案,在软件虚拟化方式下,与最新的优化方案相比,延时降低了30%,带宽有10%的提升;在使用硬件辅助虚拟化的系统中,通过FLMS能够获得接近原生系统的I/O性能,并且保证了整个系统的公平性.
[Abstract]:At present, virtualization has been widely used in data centers, but the mainstream virtual CPU scheduling policy does not guarantee the performance of I / O. Especially, when the delay sensitive load virtual machine and the computing sensitive load virtual machine compete for CPU resources, the performance of the virtual machine is significantly decreased. In view of the above problems, a flexible method is proposed. An efficient Virtual CPU scheduling algorithm. Flexible I / O latency and multi-processor sensitive scheduler. FLMS is classified by using virtual machines. Virtual CPU binding, multi-class time slice and other technologies reduce the response delay of virtual machine, and redesign the load balancing strategy based on multi-processor architecture. Under the software virtualization mode, compared with the latest optimization scheme, the delay is reduced by 30%, and the bandwidth is increased by 10%. In the hardware-assisted virtualization system, the I / O performance of the native system can be obtained by FLMS, and the fairness of the whole system can be guaranteed.
【作者单位】：武汉光电国家实验室(华中科技大学);
【基金】：国家高技术研究发展计划(863)(2015AA015301,2015AA016701) 武汉市应用基本研究计划(2015010101010004)~~
【分类号】：TP332
【正文快照】： www.jos.org.cn/1000-9825/5059.htm英文引用格式:Liu KN,Tong W,Feng D,Liu JN,Zhang J.Flexible and efficient VCPU scheduling algorithm.Ruan Jian Xue Bao/Journal of Software,2017,28(2):398?410(in Chinese).http://www.jos.org.cn/1000-9825/5059.htmFlexible and E

【相似文献】