ENHANCING THROTTLED LOAD BALANCING ALGORITHM WITH MACHINE LEARNING FOR DYNAMIC RESOURCE ALLOCATION IN CLOUD COMPUTING ENVIRONMENTS
Terence Matiwure, Arthur Ndlovu · International Journal of Computer Science and Mobile Computing · 2025
Efficient load balancing in cloud environments is critical to ensuring low response time, optimal execution time, high resource utilization, and evenly distributed server workloads. Inadequate load balancing, often due to limited integration of intelligent decision-making, can result in delayed responses and dissatisfied clients. This research proposes an intelligent enhancement to the throttled Weighted Round Robin (WRR) load balancing algorithm using Machine Learning (ML). The enhanced load balancer employs a Gradient Boosting Regressor—a tree-based machine learning model—to predict the most suitable server for each incoming task based on task profiles and current server load conditions. When the predicted server is available and offers the most efficient processing time, it is selected; otherwise, the system falls back to a WRR-based selection mechanism to identify the next best server. The entire simulation is implemented in Python and hosted using Docker Desktop (Sigmoid), providing a lightweight, reproducible environment for experimentation. Bulk tasks are submitted via CSV files and analysed by two routes: one using ML-aided WRR logic and another relying solely on pure WRR. Experimental results show that the proposed ML-enhanced approach significantly improves task distribution efficiency, reduces response and execution times, and better utilizes available server resources compared to throttled weighted round robin load balancing method.