HAVS: Heterogeneity Aware VCPU Scheduler for Virtualization Platform
Single-ISA heterogeneous multicore processors have been developed rapidly for its better performance per watt compared to homogeneous processors 2.9). With diverse workloads, heterogeneous systems can deliver more performance per watt than homogeneous ones. Because vCPUs can be mapped to cores according to the relative benefit that they derive from running on different cores with different features. In order to fully tap into this potential, the vCPU scheduler should be heterogeneity aware, so it can load vCPUs to most-matched processor cores according to the characteristics of the workload. In this paper, we proposed a heterogeneity aware vCPU scheduling algorithm which does the matching on the basis of current Xen Credit Scheduler (Xen CS). Our scheduler not only inherits the two prominent good point of CS: equity and load balance, but also gets a better performance on heterogeneous processors for its heterogeneity awareness. We implemented a prototype (HAVS) based on Xen virtual machine and made an evaluation. Test result indicates that our method achieved about 30％ performance improvement than current Xen CS when running diverse workloads on heterogeneous processors.
virtualization heterogeneous multicore scheduling power
Jie Bao Jiong Zhang Xiang Long Chengjian Wen
School of Computer Science and Technology Beihang University Beijing 100191, P.R China
国际会议
2011 International Conference on Information and Industrial Electronics(2011年信息与工业电子国际会议 ICIIE 2011)
成都
英文
842-846
2011-01-14（万方平台首次上网日期，不代表论文的发表时间）