An Efficient Tenant LED Virtual Machine Scheduling Using Machine Learning

Abstract

Cloud data centers often face the dual challenge of maintaining workload balance and maximizing resource utilization due to the highly dynamic and heterogeneous nature of hosted applications. To address these limitations, this study presents a hybrid VM scheduling framework that combines Bayes-based clustering with Particle Swarm Optimization (PSO). The model first applies clustering to group tasks based on workload similarities and resource demand characteristics, while Bayesian probability reasoning is employed to refine host selection and minimize overload risk. Subsequently, PSO is utilized to search iteratively for an optimal scheduling solution by employing a fitness function that considers response time, energy efficiency, and resource utilization. A matrix-based allocation model is introduced to represent scheduling states and guide final deployment decisions in each iteration. Experimental outcomes show that the proposed approach enhances load balancing, reduces execution delays, and improves system scalability, thereby ensuring more efficient and adaptive performance in multi-tenant cloud environments.