An Energy Efficient Virtual Machine Allocation Policy in Cloud Environment

Abstract

Data centers hosting Cloud applications consume huge amounts of electrical energy, contributing to high operational costs of the system. By supporting the movement of VMs between physical nodes, it enables dynamic migration of virtual machine according to the performance requirements and thus the nodes that are idle can be switched off to eliminate power consumption. The proper VM allocation and selection policy is required to maximize the usage of host still maintaining the service level agreements. Many allocation policies such as IQR, MAD etc are there, but such policies consider only 2/3rd of whole data set. These policies also increase the computation overhead. This thesis focused on the enhancement of virtual machine (VM) allocation policies in such a way that energy consumption, number of VM migrations and SLA violation can be reduced. Central Limit theorem in conjunction with minimum migration time selection policy has been used to achieve the expected results. Proof of correctness is derived in order to conform that the proposed algorithm is a novel approach. Validation of proposed algorithm has been done by conducting a performance evaluation study using the Cloud Sim 3.0 simulator. Testing results show that the proposed algorithm is better as compared to IQR, MAD allocation policies in terms of energy efficiency, number of VM migrations and SLA violation.