An Efficient Graph Neural Network‐Based Routing Optimization Mechanism for Congestion‐Aware Software‐Defined Networks

Software‐Defined Networking (SDN) has developed as a suitable solution for complex networks with a flexible and programmable networking paradigm. SDN infrastructure is facing critical challenges in efficient routing under dynamic and congestion‐prone conditions. This research paper proposes an efficient Improved Graph Neural Network (IGNN)‐based routing optimization mechanism designed for achieving congestion‐aware traffic management in SDN environments. The proposed model uses the inherent graph structure of network topology by representing switches as nodes and communication links as edges, and this mode is utilizing real‐time network metrics such as bandwidth, delay, queue length, and packet loss as feature inputs. A deep GNN framework is used for learning topology‐aware node by incorporating through message passing, and this will enable accurate prediction of optimal routing paths under varying traffic conditions. Unlike conventional GNN‐based routing approaches that primarily learn topology‐aware node representations through graph message passing, the proposed IGNN integrates topology representation with congestion prediction and multi‐objective routing optimization considering throughput, packet delay, packet loss, and congestion utilization, enabling congestion‐aware adaptive path selection. The suggested model makes use of congestion‐aware cost functions for low‐latency and high‐throughput routing and dynamically updates SDN flow rules using GNN predictions. The experimental evaluation conducted using a simulated SDN environment demonstrates significant performance improvements compared to traditional routing protocols and baseline machine learning approaches. The proposed method achieves approximately 23–31% reduction in end‐to‐end delay, 18–22% decrease in packet loss, and 27–33% improvement in throughput. The IGNN approach presented is a good solution to optimize SDN routing through congestion awareness.