REDO: A Reinforcement Learning-based Dynamic Routing Algorithm Selection Method for SDN

Abstract

Full bibliographic details must be given when referring to, or quoting from full items including the author's name, the title of the work, publication details where relevant (place, publisher, date), pagination, and for theses or dissertations the awarding institution, the degree type awarded, and the date of the award.

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

“…The performance of LearnSDN is compared against other state-of-the-art routing algorithms from the literature, such as: MHA, WSP, SWP, MIRA and our previous work, REDO [9]. The comparison is done in terms of throughput, packet loss rate, rejection rate, PSNR and Mean Opinion Score (MOS).…”

“…Where α f , β f , and γ f representing decision variables of value 0 if flow f satisfies the SLA in terms of throughout Θ f ,t hr , packet loss Θ f ,l oss , and rejection rate Θ f ,r e j , and 1 otherwise. The defined optimization problem would be similar to our previous work, REDO in [9]. However, the difference appears at the definition of the action space.…”

“…RED-STAR classifies the service network traffic and makes use of a differentiated reward scheme to dynamically distribute the appropriate routes to the specific service traffic. In our previous work [9], [10], we used RL to select the best routing algorithm for the QoS-based traffic flows only, while the routing strategy for the background traffic flows was set static. However, this approach would actually disturb the users' Quality of Experience (QoE) of QoS-based traffic due to constant interruptions triggered by periodical re-routing decisions.…”

Review on Machine Learning for Intelligent Routing, Key Requirement and Challenges Towards 6G

“…In [54], the author suggests a dynamic routing method selection approach that utilizes Q Learning. The suggested algorithm is trained to select the best traditional routing method to apply to the traffic flows in an SDN environment and decide which QoS-based traffic class offers the optimum balance between throughput, packet loss, and rate of rejection.…”

Optimizing Time-Sensitive Software-Defined Wireless Networks With Reinforcement Learning

“…They introduced a heuristic algorithm to resolve the problem, deriving the constraints and worst delay offering optimized bandwidth to be allocated. Garcia-Olano et al [25] applied RL for SDN, but most of them are to make optimized routing solutions of SDN nodes. Unlike these studies, we implemented TSN with SDN and proposed a resource allocation method that can be used when TSN-enabled traffic and background traffic are mixed.…”