Load_Balancing-SDN_controller

A networking architecture known as a "software-defined network" (SDN) separates the data plane of network devices from the control plane. Due to its improved centralized control and network programmability, SDN offers considerable advantages over traditional networks. To address the issue of scalability, the control plane is diversified but logically centralized. A multi-controller SDN organization needs a load balancing technique to efficiently handle local overloads since network traffic is both geographically and temporally dynamic. Building a system that learns from network traffic with the goal of swiftly balancing it, reducing needless migration costs, and obtaining a higher in-packet request response rate is the primary motivation behind this project. We use reinforcement learning to achieve global optimal controller load balancing at the lowest possible cost while maintaining the highest efficiency and avoiding migration conflicts. Our study yields better outcomes in comparison to greedy and random selection methods.