ML-Based Intelligent Control and Optimization for Enhanced BGP Path Selection in SDN

Software-Defined Networks (SDN) have emerged as a novel network paradigm that simplifies network architecture and improves utilization by modifying the structure of the traditional network system. The deployment of SDN has gradually increased in data centers and Internet Service Provider (ISP) networks for IoT traffic. The centralization and abstraction of network management over lower-level distributed forwarding engines have been provided by SDN controllers in ISP autonomous systems that employ external Border Gateway Protocol (eBGP) to perform traffic forwarding. However, static configurations and a list of attributes govern the process of selecting the best path for routing, resulting in a highly administrative configuration workload and high-risk error probability. Furthermore, these attributes are unable to manage heterogeneous dynamic circumstances, such as link congestion and dynamic load balancing. This paper proposes an intelligent combined route control and optimization approach that is based on the Genetic Algorithm (GA) and Deep Neural Networks (DNN) to overcome these issues. The proposed model is extensively tested, and the combined optimization technique efficiently identifies non-optimal routes with high congestion weights and predicts the probability of the best optimal routes. The achieved Fitness probability for generation 999 is 12.54% with prediction accuracy of 92.8%. In addition, this paper identifies the limitations of the current BGP path selection algorithm and proposes an enhanced route selection process for the global BGP routing table in the SDN control plane using an intelligent optimization approach to increase the reliability of inter-networking within autonomous systems and reduce outages in the production environment.