Q-Learning-Driven Framework for Optimized Virtual Machine Placement in Cloud Data Centers

Abstract

In modern cloud data centers, efficient virtual machine placement is crucial for optimizing resource utilization, reducing computational costs, and enhancing overall system performance. Traditional approaches, often rooted in heuristic and rule-based strategies, struggle with scalability and adapting to dynamic workloads. An innovative framework for virtual machine deployment is presented in this research, utilizing the Q-learning algorithm, a type of reinforcement learning intended to maximize decision-making under dynamic and unpredictable circumstances. Through the use of Q-learning, our framework minimizes resource contention and distributes loads among physical servers by automatically learning the best placement techniques through interactions with the cloud environment. The Q-learning strategy improves resource allocation efficiency by continuously adapting to changing situations, in contrast to static models. The suggested framework shows notable improvements in lowering makespan about 92% that is highly optimizing the resources utilization and lowering computational costs about 34%, thus framework enhancing overall system performance through comprehensive simulations and comparisons with traditional models. The outcomes demonstrate how Q-learning algorithm can improve virtual machine placement tactics and offer a flexible and scalable approach to cloud infrastructure management.