A Q-learning Based Load Balanced and QoS-aware SDN Approach: A Case Study in Defence Industry

Abstract

Traditional network routing methods are insufficient in the face of exponentially increasing data, device diversity and service demands' variety, making the necessity of more manageable routing methods felt at both the Internet and Local Area Networks (LAN). It is predicted that the Software Defined Networks (SDN) are going to be able to manage the data traffic of the digital world in a more democratic way with smart algorithms, due to their programmable and centralized management structure. In this study, the network environment of a defense industry company that has a LAN infrastructure on a large campus was selected as a case study. This internal network serves a a wide variety of devices with dozens of switches. A part of its topology and its traffic that currently routed with Spanning Tree Protocol (STP) has been simulated with widely used methods. In this network environment; various test scenarios have been studied with STP, a rule-based SDN, and our Q-learning based SDN approach. Our proposal, which performs the load balancing of the system while providing the requested QoS to the clients, has achieved effective results under various QoS performance metrics and load balancing indicators.