QoS-Oriented Adaptive Routing in Distributed SDN using SARSA Reinforcement Learning

Software-Defined Networking (SDN) is an architectural approach that offers a more adaptable, programmable method to network management by splitting the control and data planes. Despite efficiently using network resources, SDN can cause congestion, longer end-to-end delays, and overburdened links. To tackle these issues, this study uses distributed SDN controllers that spread the control logic among multiple nodes, improving scalability. This research utilizes the State Action Reward State Action (SARSA) reinforcement learning technique. This method dynamically learns and adapts routing decisions based on Quality of Service (QoS) values, which helps make intelligent routing decisions. With the inclusion of QoS parameters, distributed control logic, and smart routing algorithms, the aim is to solve a path optimization problem constrained by QoS requirements. The evaluation of QoS performance is achieved through a reward function of SARSA that tries to find an optimal sequence with the minimum number of intermediate hops between the source and destination. This points to the success of the SARSA algorithm in distinguishing between various traffic and choosing the best routes based on QoS requirements. The study provides beneficial insights into the practicality and efficiency of the suggested architecture in various network environments. It significantly advances SDN solutions, guiding future research and development to optimize network performance and QoS in diverse, modern networking scenarios such as those required for 5G networks and the beyond.