Load Balancing in SDN-Based Cloud Infrastructures

Abstract

The convergence of Software-Defined Networking (SDN) and cloud computing has transformed infrastructure management by decoupling the control and data planes, offering centralized programmability, increased agility, and scalability for cloud environments. Yet, efficient load balancing—ensuring optimal distribution of traffic and processing across controllers and servers—remains a persistent challenge. This paper surveys mechanisms and frameworks for SDN-enabled cloud load balancing, including multi-controller strategies, reinforcement learning (RL)- based switch migrations, and adaptive algorithmic approaches. Key techniques are analyzed in terms of response time, throughput, packet loss, and scalability. Moreover, we propose a hybrid methodology combining centralized controller oversight with dynamic RL-driven offloading and switch reassignments to achieve both efficiency and resilience in cloud data centers. Empirical evaluations from literature reveal that load balancing techniques like Efficiency-Aware Switch Migration (EASM) can reduce controller response time by ~22% and increase throughput by ~30%. Other strategies using modified Bully algorithms demonstrate significant improvements in packet transmission ratio and reduced packet loss. A systematic workflow involving monitoring, load detection, decision-making, load adjustment, and performance feedback is outlined to guide SDN-based implementations. Advantages such as improved scalability, high throughput, and dynamic adaptability are weighed against challenges like migration overhead, algorithm complexity, and requirement for high-quality telemetry. The paper concludes with insights on deploying RL-based and multi-controller frameworks effectively and outlines future research avenues including predictive balancing with timeseries forecasting, energy-aware scheduling, federated controllers, and AI-driven decision explainability—all predating 2022 literature.