Application performance analysis during live migration of virtual machines

Abstract

Live migration is an essential feature in virtualization technology where a running Virtual Machine (VM) from one physical host is migrated to another physical host without any service disruptions. Indisputably the benefits reaped from VM migration are high availability, load balancing, energy saving and disaster recovery which are the desired data centre attributes. The migration is initiated by the administrator and part of the migration procedure is to make an informed decision to isolate and identify an appropriate VM to be migrated, lest an impressive performance may not be achieved. The decision of selecting the right candidate VM to be migrated depends on parameters like total migration time, down time, total transferred data and page dirty rate. The nature of the application influences these parameters. The paper addresses the analysis of some of these parameters empirically, and based on these data a correct VM to be migrated is chosen. In addition to the above, this paper also discusses the dynamic resource allocation for RSA algorithm and JMeter.