Simulated Annealing based approach for Virtual Machine Live Migration

Abstract

Virtual machine migration technique is used in cloud computing to increase the reliability and scalability of the cloud computing systems. It helps the service providers to achieve resource efficiency and quality of service. At the time of live migration, the underlying virtual machine continuously works until the entire or part of data is migrated from source to destination. Live migration of Virtual Machine (VM) acts as an important technique that allows the management of resources, maintenance of server and load balancing in cloud data centers, but there is a degradation of performance at the source and destination physical machines. Different live migrations techniques have been proposed, each showing different properties like completion time, amount of data transferred, down time of VM and degradation of performance of VM. In this paper, a live migration algorithm, Live Migration Annealing (LMA) Virtual Machine Migration that makes use of an evaluation function to perform analysis on the time series data collected over the iterations made during the live migration period is proposed. It embodies the concept of exploration and exploitation of knowledge and space. It also takes ideas from the Iterative depth first Search to perform the iterations and Simulated Annealing to find the pages eligible for the live Migration. All the eligible pages undergo a phase called Selection phase before being finally sent to the destination virtual machine which incorporates the idea of Second Chance. Live Migration is only performed if it isn’t happening at the expense of the downtime. An overall decrease in the number of iterations and downtime with least possible live migration time is achieved through the proposed algorithm.