With their rigid layers, traditional networks do not meet evolving traffic demands. As a result, they tend to face congestion along with un-optimized routing. SDN controls traffic management by introducing a programmable control plane, enabling dynamic and intelligent network management. However, older routing techniques, such as Dijkstra's and Multipath, suffer from low adaptability, leading to a rise in latency and packet loss. The addition of Q-learning with Q-Optimizer in SDN is the aim of this study in order to improve the Quality-of-Service metrics, such as throughput, Round Trip Time (RTT), jitter, and Packet Loss Ratio (PLR). Experimental results from Mininet using the Ryu controller demonstrate that Q-Optimizer improves throughput by 36.49%, reduces RTT by 46.09%, minimizes jitter by 95.01%, and lowers Packet Loss Ratio (PLR) by 63.32% compared to Dijkstra’s algorithm. Compared to Multipath routing, Q-Optimizer improves throughput by 13.25%, reduces RTT by 33.22%, decreases jitter by 25.32%, and lowers PLR by 55.61%. Even compared to Q-Learning, it shows improvements in achieving an 11.76% increase in throughput, 26.05% lower RTT, 14.81% less jitter, and 34.48% lower PLR. The statistical validation using one-way ANOVA confirms that these improvements are significant, reinforcing Q-Optimizer's effectiveness in SDN environments. A one-way ANOVA test (F = 785.78, p = 0.0000). The outcomes reveal that AI-driven SDN frameworks are more impactful than traditional approaches and provide scalable and innovative solutions to current global networking infrastructures.
Goteti et al. (2026) studied this question.