Modern cloud computing and distributed networks face unprecedented traffic volatility, rendering traditional, static load-balancing algorithms—such as Round Robin or Least Connections—increasingly inefficient. Intelligent load balancing, driven by machine learning (ML), has emerged as a transformative solution to manage these dynamic workloads. By leveraging historical data and real-time metrics, ML models can predict traffic surges, identify resource bottlenecks, and autonomously redistribute tasks to optimize Quality of Service (QoS). This review explores the paradigm shift from reactive to proactive traffic management. We examine various ML architectures, including supervised learning for resource estimation, unsupervised clustering for traffic classification, and reinforcement learning for real-time decision-making. The article synthesizes current research on multi-objective optimization, focusing on the trade-offs between energy efficiency, latency reduction, and throughput maximization. Finally, we discuss the challenges of implementing these models in edge and fog computing environments, providing a roadmap for future developments in self-healing, autonomous network infrastructures.
Javlon Ismailov (Wed,) studied this question.