Virtual Machine Migration Methods for Heterogeneous Power Consumption

Virtualization is widely used as a basis for cloud computing. For this technique, if self-management of virtual machines (VMs) can be developed, operational costs will be greatly reduced. To develop self-managed virtualization, it is necessary to develop an algorithm that optimally places VMs on physical machines (PMs) and executes VM migrations among PMs in response to changes in demand. The algorithm must satisfy various requirements, i.e., Minimum power consumption, sufficiently good performance, relatively few migrations, and quick computation. This study proposes three different algorithms for executing VM migrations in order to save electrical power. The proposed methods are based on a greedy algorithm and employ different ways of searching for PMs and VMs involved in migrations. The methods employ an efficiency metric defined in terms of resource usage and electric power for an environment in which power consumption is heterogeneous among PMs. The proposed methods are evaluated via computer simulations. Among these methods, we find there is a trade off between power consumption and the number of migrations. We also find that the most power conserving method achieves power consumption that is close to the strict minimum power.