Asymmetric Virtual Machine Scheduling Model Based on Workload Classification

Multi-core virtualization platform is an essential basement of modern computing environment including server consolidation and cloud computing. However hidden contention for physical resources between different workloads always causes serious performance interference and thus reduces system performance and efficiency. In order to address this issue, we exploit the possibility of scheduling virtual machine asymmetrically based on workload classification and multi-core partitioning. In this model, available processor cores of system are divided into several individual subsets which employ specific scheduling algorithms to undertake CPU-intensive and I/O-intensive missions respectively, and different optimizations are applied based on workload characteristic. We implement a prototype based on Xen virtual machine and preliminary test results demonstrate that this scheduling model can efficiently reduce the performance interference caused by scheduling competition in multi-core virtual machine system.