Network Failure and Anomaly Prediction to Achieve Quality of Service (QoS) on Software-Defined Networks

Abstract

We develop a predictive optimization program to resolve anomalies and failures on Software Defined Networks (SDN) proactively in order to prevent such failures before they render important services like health security and production unavailable. The previous studies on preventing network anomalies or failures took a reactive approach by which the anomalies are resolved after they occur. Our program predicts if the incoming 5G flows will cause an anomaly on the nodes by using machine learning and then leverages a linear optimization program to find the best routes for such flows to be admitted safely. Our program is network topology agnostic, hence it can be run on any topology. Since our approach resolves such anomalies proactively and makes sure the important services are always continuous and available for the communities it holds the potential to impact the design of SDNs in the near future. © 2022 Elsevier B.V. All rights reserved.