Energy Efficient Technique for Live Virtual Machine Migration

Abstract

Cloud is dynamic platform for delivering computing services. The demand of Cloud infrastructure has rapidly expanded with an increasing data rate of 50% every year. The increased volume of data results in more processors, larger storage devices and more administration efforts. It is estimated that in next ten years the data consumed by datacenter will be more than 150 Terawatt hours (TWh). The percentage of data consumption and cooling equipment will reach 10% of the total energy consumption in world. Virtualization is the technique widely used in modern data-centers in order to realize energy efficient operations. Live Virtual Machine (VM) migration plays an important role in Cloud and holds various benefits such as energy aware server consolidation, load balancing and resource distribution. It causes transfer of large unnecessary memory pages known as dirty pages and thus leads to increase in downtime and total migration time. Existing approach of live VM migration suffers from problems like high energy consumption, high response time and more migrations. Proposed approach of optimizing live VM migration by using hybrid BFO (Bacterial Foraging Optimization) algorithm reduces unnecessary migrations and execute all Cloudlets in minimum time and least migrations. BFO algorithm divides workload over VM network based upon the performance of nodes. Then migration is performed on the network with the help of Post- Copy technique. This architecture makes the execution process highly accurate and less time-consuming. VM allocation and load balancing is done in which processing of VM load is checked along with length of each Cloudlet. The optimal host is selected for migration through BFO algorithm and Post-Copy algorithm is applied which migrate the VM and launch migration process without interrupting the execution. This thesis focuses on providing energy efficient Live VM migration technique so that number of migrations can be reduced and Cloudlets can finish their tasks efficiently in minimum time with less consumption of energy.