Analysis of SD-WAN Packets using Machine Learning Algorithm

Abstract

In recent years, legacy networks have evolved to incorporate the use of programmability features with the aim of improving performance and resource utilisation. In achieving this goal, packets need to be monitored and classified. In this study, an optimal monitoring tool is used in capturing the packets or flows in an emulated Software Defined Wide Area Network using GNS3. The network architecture is implemented using two hosts communicating to a server integrated with a machine learning (ML) model (python based) to classify real network packets. The ML model is achieved using the Decision Tree algorithm based on python programming. The proposed implementation ensures the ML algorithm efficiently classifies and segments various packets in the network in a database structure. This testbed can be effectively implemented in a real network scenario, and packet data can be captured and analysed into a database structure which can be used for further analysis such as congestion window or throughput for improving network performance and resource utilisation.