Traffic Data Classification using Machine Learning Algorithms in SDN Networks
Jungmin Kwon, Daeun Jung, Hyunggon Park
Abstract
As an efficient approach to proactively monitoring network dynamics, automatically analyzing network data, and predicting network usage, machine learning has been widely deployed. This enables the networks to be efficiently and autonomously coped with in SDN/NFV environment. In particular, network intelligent technology can be adopted into the infrastructure management, network operations, and service assurance. In this paper, we study the automatic network data classification based on machine learning, where several machine learning algorithms are deployed to automatically classify real network traffic data collected from ONOS (Open Network Operating System) platform. From the experiment results with simple network topology, we conclude that machine learning algorithms can effectively classify the network traffic data. However, it is also observed machine algorithms may only show a limited performance in practice if they are blindly deployed. This is because there exists not only the data that needs to be delivered to the receivers but also the data required for network maintenance in a real network system. Therefore, it is essential to develop machine learning algorithms that explicitly consider the characteristics of real network traffic data in target network scenarios.