LearnSDN: Optimizing Routing Over Multimedia-based 5G-SDN using Machine Learning

Abstract

With the advent of 5G networks and beyond, there is an increasing demand to leverage Machine Learning (ML) capabilities and develop new and innovative solutions that could achieve efficient use of network resources and improve users' Quality of Experience (QoE). One of the key enabling technologies for 5G networks is Software Defined Networking (SDN) as it enables fine-grained monitoring and control of the network. Given the variety of dynamic networking conditions within 5G-SDN environments and the diversity of routing algorithms, an intelligent control of these strategies should exist to maximize the Quality of Service (QoS) provisioning of multimedia traffic with more stringent requirements without penalizing the performance of the background traffic. This paper proposes LearnSDN, an innovative ML-based solution that enables QoS provisioning over multimedia-based 5G-SDN environments. LearnSDN uses ML to learn the most convenient routing algorithm to be employed on the background traffic based on the dynamic network conditions in order to cater for the QoS requirements of the multimedia traffic. The performance of the proposed LearnSDN solution is evaluated under a realistic emulation-based SDN environment. The results indicate that LearnSDN outperforms other state-of-the-art solutions in terms of QoS provisioning, PSNR and MOS.