An Extended Energy-Aware Cost Recovery Approach for Virtual Machine Migration

Energy supply problems, fuel costs, and global warming all suggest the continued need to explore possible savings through efficient scheduling and consolidation techniques. However, consolidation involves migrations that cost additional energy, so it can be more energy efficient not to consolidate. In several experiments, we found that efficient resource allocation methods can be more energy efficient than migration-based techniques, where it is possible to avoid inefficient migrations, i.e., those that can never recover the energy required for migration. In this paper, we elaborate our previous results and demonstrate through extensive experiments, using the Google workload traces for 12 583 hosts and 20 071 200 tasks, how different virtual machine (VM) allocation heuristics, combined with different migration techniques, will impact on energy efficiency. We suggest, using reasonable assumptions for datacenter setup, that a combination of energy-aware Fill-Up VM allocation and migration, and migration only for relatively long-running VMs, provides optimal energy efficiency. Furthermore, switching off the idle hosts provides more energy efficiency, however, at a nonnegligible cost in terms of scheduling delay.