An Innovative Reinforcement Learning-Based Framework for Quality of Service Provisioning Over Multimedia-Based SDN Environments

Within the current global context, the coronavirus pandemic has led to an unprecedented surge in the Internet traffic, with most of the traffic represented by video. The improved wired and guided network infrastructure along with the emerging 5G networks enables the provisioning of increased bandwidth support while the virtualization introduced by the integration of Software Defined Networks (SDN) enables traffic management and remote orchestration of networking devices. However, the popularity and variety of multimedia-rich applications along with the increased number of users has led to an ever increasing pressure that these multimedia-rich content applications are placing on the underlying networks. Consequently, a simple increase in the system capacity will not be enough and an intelligent traffic management solution is required to enable the Quality of Service (QoS) provisioning. In this context, this paper proposes a Reinforcement Learning (RL)-based framework within a multimedia-based SDN environment, that decides on the most suitable routing algorithm to be applied on the QoS-based traffic flows to improve QoS provisioning. The proposed RL-based solution was implemented and evaluated using an experimental setup under a realistic SDN environment and compared against other state-of-the-art solutions from the literature in terms of throughput, packet loss, latency, peak signal-to-noise ratio (PSNR) and mean opinion score (MOS). The proposed RL-based framework finds the best trade-off between QoS vs. Quality of User Experience (QoE) when compared to other state-of-the-art approaches.