Abstract

The importance of the core internet infrastructure’s scalability, resilience, and adaptability has grown tremendously due to shifting traffic patterns, diverse network requirements, increasing security concerns, and dynamic flow of data. Although traditional routing techniques are quite effective and proven robust over time, they do not adaptively respond to real-time traffic changes and state of the network. In this paper we present ASDN-ARA (Adaptive Software-Defined Networking-based Routing Architecture), a new model aimed at improving responsiveness alongside performance with regard to core routing by using SDN programmability. ASDN-ARA presents a new approach to adaptive routing with its multi-layered architecture composed of telemetry streams in real time, intent-driven policy engines, as well as modules of traffic engineering which are aware of network topologies. The architecture shifts its control over the decision layer based on congestion per link and latency thresholds, route flapping detection, as well as application-level QoS metrics. The ASDN-ARA controller is built around a hybrid optimization algorithm that employs Reinforcement Learning (RL) combined with heuristic path scoring to find optimal forwarding paths between transit AS networks. Simulation and emulation tests performed on GNS3-based testbeds show core findings that these proposed mechanisms can achieve up to 28% reduction in average end-to-end latency while increasing link utilization balance by 35%. Furthermore ASDN-ARA outperformed conventional OSPF and BGP based routing schemes in lowering convergence time during link failures. The system shows compliance with policies in multi-domain environments utilizing segment routing extensions. In addition, ASDN-ARA maintains integration compatibility with legacy control planes through an adaptive translation module, which allows for progressive implementation within the frameworks of pre-existing ISP infrastructures. The introduced adaptive routing model enhances the intelligent core routing paradigm by allowing greater autonomy in policy adherence, automation as well as latency-sensitive data packet transmission across extensive networks.