A Methodology for Redistributing the Operating Virtual Machines among the Servers in a Data Center

Abstract

The need for centralized processing and storage of data is growing every year. Data centers of still greater capacity and computing power are put into operation for this purpose. To reduce their expenses the owners of data centers pay special attention to optimizing the energy consumed by the data center infrastructure. An approach for improving the data center’s overall energy efficiency through redistributing its load with subsequently shutting down the idle servers is considered. A mechanism governing the migration of virtual machines for optimizing power consumption by the servers is outlined. The existing methods used to redistribute virtual machines either perform a full search which is impossible in the case of real dimensions or disregard the power consumption criterion. A procedure is proposed that excludes the need to perform a full search and places focus on improving the efficiency according to the criterion of total energy consumption by the data center servers due to redistribution of the operating virtual machines among the data center servers. The proposed method is based on a heuristic procedure of selecting the destination servers and consists of a static part, which collects information about the available resources, and a dynamic part in which the virtual machines are distributed among the servers depending on the initial data and the preset limiting parameters. A model example demonstrating the operability of the method is given which outlines the steps in applying the method and shows how the destination servers are selected for migrations.