Load Balancing Optimization in Cloud-Based Distributed Systems using Machine Learning

Abstract

Cloud-based distributed systems are made up of a number of interconnected servers, virtual machines, containers, and computing resources which handle large volumes of user requests at the same time. It is necessary to have efficient load balancing to keep the system performing well, reduce response time, improve the use of resources, and avoid server overload. Common load-balancing methods such as Round Robin, Least Connections, and Weighted Load Balancing usually rely on fixed rules and are not able to respond effectively to workloads that change frequently. The study presents a framework for optimising load balancing in cloud-based distributed systems using machine learning. This approach gathers both real-time and historical information such as CPU usage, memory usage, network traffic, response time, the number of active requests, and server capacity. Machine learning models are then applied in order to predict the workload situation and to choose appropriate servers for the incoming requests. The framework examines the use of models such as Random Forest, Gradient Boosting, Artificial Neural Network (ANN) models and compares their performance against traditional load-balancing algorithms. The system is assessed in terms of response time, throughput, resource utilisation, load imbalance, SLA violation rate, and prediction accuracy. The anticipated result is an adaptive load-balancing system which is able to distribute the workloads more efficiently while at the same time reducing response time and server overload..