Load Balancer Decision System: A Reinforcement Learning Approach for Optimizing Distributed Computing Environments

In this work, we introduce a reinforcement learning framework to tackle load balancing challenges in distributed computing platforms. Moving beyond conventional static schedulers like Round Robin, our approach intelligently routes tasks to servers by continuously monitoring critical metrics—task volume, CPU/memory utilization, network latency, and execution priority. At the core of our system is an interactive GUI through which users can monitor live task distribution, process batch workloads, train the Q-Learning model,visualize real-time performance metrics, and compare results with baseline scheduling algorithms. Built with modularity in mind, the architecture ensures seamless scalability and upkeep through specialized modules: a core Q-Learning agent and an advanced visualization pipeline. Fundamentally, our solution discretizes the state space, maintains adaptive Q-tables, and crafts sophisticated reward signals to drive optimal task placement decisions. That said, we acknowledge key limitations: The training pipeline remains partially developed, algorithm comparisons rely on simulated data, and Q-table sizes are currently static. Moving forward, our roadmap includes a complete Q-Learning implementation, empirical validation across diverse workloads, integration of Deep Q-Networks for scalable state representation, and support for real-time data feeds to boost responsiveness.