Efficient Virtual Machine Scheduling-policy for Virtualized Heterogeneous Multicore Systems

Heterogeneous multicore processors could be the future trend of processors’ industry due to their performance-power efficiency. In the operating systems domain, A heterogeneity-aware scheduler assigns a thread or an application to an appropriate core to realize this efficiency. Using virtualization technologies enables resource consolidation and achieves effective utilization of resources. Nevertheless, Hypervisors’ scheduling-policy is based on the round robin algorithm to ensure fairness among VMs. Emerging heterogeneous system and virtualization increases power savings and enhances resources utilization. This combination needs a new scheduler, which schedules each VM to an appropriate core based on its characteristics. In this paper, we present sources of delay in virtualized environment that could degrade performance of VMs. Then, we investigate the sensitivity of a VM performance to changes in clock frequency. A new scheduling policy was implemented to alleviate sources of delay and to be aware of system’s heterogeneity. We emulated our heterogeneous testing environment using DVFS, and we compared the results to default scheduling policy in Hypervisor’s scheduler. Nevertheless, the results show performance improvements for VMs that run either CPU-intensive or I/O-intensive applications. Finally, the measured power savings of our heterogeneous testing environment reach up to 25%.