Optimal VM Placement Approach Using Fuzzy Reinforcement Learning for Cloud Data Centers

Abstract

Ineffective resource management in cloud data centers leads to high energy usage and cost to maintain the resources. A Virtual machine (VM) placement is a part of the VM consolidation problem where each VM should be mapped to the available host. An optimal VM placement is an effective way to enhance both resource usage and energy efficiency. However, minimizing the energy consumption without affecting SLA violation and performance of the application is quite challenging. This paper proposes a fuzzy based State-Action-Reward-State-Action (SARSA) reinforcement learning algorithm to address the VM placement problem. The integration of fuzzy controller with Reinforcement learning (RL) algorithm is capable of optimizing the reallocation of maximal number of VMs into a minimal number of hosts that reduce energy usage and wastage of resource as well. The proposed work is capable of handling the fluctuating workload situations and delivers proper placement of VMs (initial or re-mapping) while ensuring the desired Quality-of-Service (QoS) demands of users to meet the Service Level Agreement (SLA). The experimental results exhibit reduced consumption of energy and less resource wastage as compared with other VM placement algorithms.