Performance Evaluation of Supervised ML Algorithms for Elephant Flow Detection in SDN

Abstract

Software-defined networking (SDN) improves the network management due to the separation of the network control plane from the packet forwarding plane. However, with the increase in data traffic, SDN architectures have raised several challenges in terms of traffic engineering, QoS, and network management. Therefore, it is crucial to develop an intelligent system to classify the flows and predict future traffic. Indeed, in order to propose an adequate forwarding strategy for various flow types (particularly elephant flows (EFs)) in an SDN environment, an accurate flow detection system is required. Hence, in this paper, we propose a model-based SDN controller that includes machine learning algorithms to detect large-size traffic and forward it. Moreover, we represent a comparative simulation to evaluate the performance of some supervised machine learning algorithms such as Naive Bayes (NB), K-Nearest neighbors (KNN), Logistics regression (RL), Support Vector Machine (SVM), and Decision Tree (DT), to detect the elephant flow. A decision tree (DT) and K-Nearest neighbors (KNN) are the best candidate machine learning algorithms in elephant flow detection with an accuracy of 99%.