Efficient VM Selection Heuristics for Dynamic VM Consolidation in Cloud Datacenters

Dynamic Consolidation of Virtual Machines (VMs) in a cloud data center requires live migration of VMs from over-utilized hosts. A primary but relatively an ignored part of consolidation process is the efficient selection of Virtual Machines from an over-utilized host for migration. Two VM selection policies, Threshold Based Selection (TBS) and Capacity Based Selection (CBS), have been proposed in this paper. These policies are based on the idea of simultaneously minimizing multiple factors that contribute to the degradation of the quality of service due to consolidation. Three degrading factors considered in the policies are, the time duration of VM migrations, time duration hosts remain over-utilized and the total number of migrations required for consolidation. Cost functions, involving these degrading factors, have been provided which formed the bases for TBS and CBS. TBS is an efficient VM selection mechanism that focuses more on the time duration of VM migrations and the total number of migrations required for the consolidation process as a trade-off between the three degrading factors. On the other hand, CBS is another efficient mechanism with more emphasis on reducing the time duration for which hosts remain over-utilized. Experiment results obtained by using Cloudsim simulating toolkit have shown that our proposed policies outperformed conventional VM selection policies like MMT, MU, and RC on indicators such as energy consumption, SLA violations, and overall performance efficiency.