ReCAP: Reliability–Capacity Aware Joint Controller Placement and Routing Using a Hybrid AI Approach

When deploying software defined network (SDN), there is a need for optimized controller placement to ensure high network performance. Different solutions are proposed that focus on single optimization objectives, such as delay minimization or throughput maximization. This article proposes a hybrid artificial intelligence-based approach, the reliability–capacity aware joint controller placement and routing (ReCAP), which combines deep reinforcement learning and graph neural networks to identify the best placement for controllers, so there are optimal paths for communication between switches and controllers and improved network performance. Controller placement is optimized based on multiple objectives, including load balancing, path reliability and link capacity. ReCAP is validated using real network traces and evaluated using the Panopticon network. The experimental results show an improved performance for our ReCAP approach compared to existing solutions in terms of both efficiency and reliability.