Autonomic and energy-aware resource allocation for efficient management of cloud data centre

Abstract

Server Virtualization is the key technology used in cloud data centers. In this technique, number of Virtual Machines (VM) can simultaneously run on the top of a single Physical Machine (PM) or host server. Each VM hosts guest operating system, middleware software and applications. There are various dimensions of resources available in PM such as CPU cores, memory, network bandwidth and storage space. As per the requirement of applications deployed, VM allocates resources from pool of resources available in a PM or host server. The placement of VMs into appropriate PM and as the need arises migrate them among other PMs by achieving application performance and saving energy are the key issues of this research paper. The performance of applications is improved by reducing frequency of live VM migrations among PMs and energy is saved by minimizing number of active servers in a data centre. The proposed approach presents algorithm for resource allocation in cloud data centre by considering various factors such as major resource requirement during initial setup of virtual machines, dynamic resource allocation at peak load on applications, performance of applications and power saving in a data centre. A data centre is simulated with heterogeneous servers by assigning load of randomly virtual machines containing CPU as well as memory intensive applications. The power consumption and VM placement failure rate are considered as parameters for analyzing the proposed algorithm. The experimental results of proposed algorithm for initial placement of VMs are compared with various algorithms such as first fit, best fit and random selection of PMs. In addition to the initial placement of VM in appropriate PM, the research issue of dynamic resource management in a data centre is also addressed.