ONU: A Reinforcement Load Balancing Approach for Resource-Aware Task Allocation in Distributed Systems

Distributed systems have revolutionized modern computing by enabling the parallel execution of large-scale tasks across interconnected nodes. However, the heterogeneity of resources in distributed systems often leads to load imbalances and underutilization of certain nodes, resulting in suboptimal performance. To address these challenges, this research paper presents a novel approach for resource-aware task distribution based on reinforcement load balancing in distributed systems with over utilized, normally utilized, and underutilized nodes. The proposed approach leverages reinforcement learning algorithms to dynamically distribute tasks among nodes based on their resource utilization levels. ONU module is designed to cater to the specific needs of each node category: the ONU Algorithm focuses on mitigating overburdened nodes, moderate resource usage nodes, and the underutilized nodes operating below capacity. The results demonstrate significant improvements in load balancing and resource utilization compared to traditional load balancing techniques.