An Energy Efficient Virtual Machine Migration Policy in Cloud Environment

Abstract

Due to rapid growth in cloud computing users, managing datacenter resources in an energy efficient way has become a key challenge. To make datacenter energy efficient virtual resources are created on the single physical hardware to fulfill the demands of large number of users simultaneously by using lesser number of active physical hosts. In virtualization, virtual machines are created on single host and these are migrated between different hosts according to current load and resource requirement for VM which results in minimizing the number of active hosts by powering off the idle hosts. Number of virtual machine migration techniques are present such as PSO, MPSO etc, which tries to minimize the energy consumption. Theses approaches minimizes the number of active servers by migration which sometimes results in overloading of remaining active servers and high energy consumption. This thesis work proposes a new virtual machine migration technique which tries to minimize the energy consumption by migrating the load between Virtual machines and running the tasks on the VMs according to VM’s load in iterative manner. Cross Pollination Optimization technique is applied to achieve the desired outcome. The proposed algorithm is validated using cloudsim 3.0 simulator with integrated workflowSim. The results from complete validation study proved that Cross Pollination Optimization performs better then Modified Particle Swarm Optimization in terms of energy consumption corresponding to number of virtual machines and CPU utilization.