SLA-Aware Load Balancing In Cloud Computing Using Machine Learning Based Virtual Machine Scheduling

Abstract

Cloud computing has become an essential platform for delivering scalable and on-demand services, but it faces significant challenges in managing resource allocation while ensuring adherence to Service Level Agreements (SLAs). This research proposes a novel SLA-aware load balancing algorithm enhanced by machine learning (ML) for efficient task scheduling across virtual machines (VMs). The proposed method predicts VM performance using real-time system metrics and allocates incoming tasks based on their SLA requirements, such as response time and priority. By integrating predictive modeling into the scheduling process, the system minimizes task failure rates, optimizes resource utilization, and maintains high SLA compliance. A simulation environment was developed using Django and a relational database backend to evaluate the algorithm against traditional methods like Round Robin, Least Connection, and Random Allocation. Experimental results showed that the ML-based approach achieved a 96.4% SLA compliance rate, significantly outperforming Round Robin (78.2%), Least Connection (84.7%), and Random Allocation (73.5%). Additionally, it reduced the average task completion time to 238 ms, compared to 412 ms, 367 ms, and 489 ms, respectively. The proposed system also improved VM utilization efficiency to 91.2%, minimized the task failure rate to 1.7%, and achieved a high prediction accuracy of 94.8%. These results confirm the effectiveness of integrating SLA constraints with intelligent ML-based scheduling to ensure performance reliability, resource efficiency, and scalability in dynamic cloud environments. This research highlights the potential of combining SLA policies with intelligent scheduling techniques to create adaptive, scalable, and performance-driven cloud infrastructures.