Fairness scheduler for virtual machines on heterogonous multi-core platforms

Abstract

Heterogeneous multi-core processors are now widely deployed to meet computation requirements for multimedia applications on embedded mobile devices. However, due to the difference on computation capability of heterogeneous multi-cores, it is challenging to share the load among cores and to better utilize the cores on the processor. In this work, we develop a fairness scheduler to share the load among cores to meet the above challenge. The developed framework ensures that each virtual machine receives its proportional share of computation time over different processing elements. To fairly schedule the tasks on the platform, VM-aware fair scheduler (VMAFS) takes into account the preemptibility of processes on co-processors. A family of algorithms is designed to schedule tasks on preemptive and non-preemptive processing elements. We define fairness on such architectures and use this metric to efficiently drive VMAFS algorithm so as to fairly manage non-preemptive computing resources. Performance evaluations result show that when VMAFS algorithm is used, scheduling fairness is not sensitive to the amount of computation time of non-preemptive tasks. In addition, VMAFS algorithm greatly outperforms credit-based scheduling algorithm, which is deployed in existing virtualization environment.