Q-Learning based VM Consolidation Approach for Enhancing Cloud Data Centres Power Efficiency

¹ Department of CSE(AI&ML), Vardhaman College of Engineering, Hyderabad, India
² Department of CSE(AI&ML), Vardhaman College of Engineering, Hyderabad, India
³ Department of CSE(AI&ML), Vardhaman College of Engineering, Hyderabad, India
⁴ Department of CSE(AI&ML), Vardhaman College of Engineering, Hyderabad, India
⁵ Department of CSE(AI&ML), Vardhaman College of Engineering, Hyderabad, India
⁶ Department of CSE(AI&ML), Vardhaman College of Engineering, Hyderabad, India

^* Corresponding author: sreenithyabaikani@gmail.com

Abstract

Energy consumption has become a common problem since days. Addressing the energy related problem is a challenging task. There are various strategies present to minimize this problem. One among them is using cloud computing infrastructure and VM setup. Virtual Machine consolidation is a viable solution to mitigate energy related issues of data centres. In recent times, we have seen various learning approaches which are used in managing the cloud data resources well. Among the approaches, Virtual Machine consolidation technique gives the viable solution for energy related issues by mitigating them. We have also delved with reinforcement learning algorithm to tackle the virtual machines. In this implementation we make use of different RL algorithms such as SARSA, Q-learning etc. and finds out the best suited algorithm. Furtherly, we will execute the model on using the algorithm chosen to build the model. The inputs we take are VM numbers, power utilization, scalability of VMs, CPU utilization time etc. and finds out what percentage of these values we are getting as an output which highlights the effectiveness of our approach, improvement in energy efficiency and service reduction etc.