Adaptive Routing and Self-Healing for Fault-Tolerant SDN Mesh Networks

Abstract

This paper presents AdaptiveNet SDN, a software-defined networking simulation framework designed to maintain communication when links, nodes, or groups of adjacent links fail. The system combines a multi-objective adaptive routing algorithm with a self-healing controller, heartbeat-based failure detection, bandwidth-aware admission control, and real UDP packet forwarding among virtual mesh nodes. Routing cost considers latency, reliability, energy use, and deadline miss probability, while an exponentially weighted moving average adjusts routing priorities according to observed network conditions. An eight-percent path retention rule reduces unnecessary route changes during small metric fluctuations. The framework was evaluated on six-node and twelve-node topologies using repeated failure experiments and a Dijkstra latencybased baseline. Across the current eighteen-condition experiment sweep, the twelve-node topology achieved complete recovery for single link, two independent link, and sequential failure scenarios for both algorithms. Node and correlated storm failures exposed the limits of topology redundancy. The experiments did not establish a statistically distinguishable advantage for adaptive routing on service recovery time or packet loss, which is reported as a limitation rather than hidden. The main contribution is therefore an integrated platform for studying adaptive routing and self-healing behavior under controlled network failures.