Game Theory Based Energy-aware Virtual Machine Placement towards Improving Resource Efficiency in Homogeneous Cloud Data Center

Abstract

In recent years, cloud data centers (CDCs) have grown into an essential component of enterprise's computing infrastructure. However, the upswing in the demand for cloud-based services is accompanied by the consumption of a considerable amount of energy. Thus, developing efficient management approaches to reduce energy consumption in CDCs becomes a top priority. Although a number of methods have already been developed for lowering the energy usage of CDCs, but they largely focus on cutting down the number of active physical machine (PM) and hardly address the issue of resource utilization efficiency in a multi-resource setting. In a dynamic environment, which often leads to frequent system overloading and sometimes causing a violation of service level agreement (SLA) in terms of quality of service (QoS) degradation. In order to tackle all the problems, a novel game theoretic approach, namely GB-VMP has been proposed in this paper to place the virtual machines (VMs) onto the PMs by leveraging the maximization of utilization across all the resource dimensions in an uniform way. The game finds the optimal mapping of VM to PM that requires minimal PMs to be in active state. It enables each of the PMs to be accommodated by a large number of VMs without causing any system overload. As a result, an energy efficient solution has been delivered while utilizing all the resources in a best possible way. Extensive experiments have been carried out on a simulated cloud setup to evaluate the effectiveness of the proposed approach in comparison to two well-known VMP strategies, IRB-VMP and BFD. The experimental results illustrates that GB- VMP outperforms the other strategies in terms of energy consumption and resource efficiency.