DeepQoSR: A Deep Reinforcement Learning based QoS-Aware Routing for Software Defined Data Center Networks

Abstract

Software Defined Networking (SDN) is an emerging technology and the fundamental principle used in the data center networks. It is expected to grow exponentially due to the addition of new IoT-enabled devices. SDN-based centralised control allows for optimal and sometimes sub-optimal network management and this can be improved by introducing Machine Intelligence to route and handle majority of the traffic. Intelligent routing is gaining popularity with the advent of Reinforcement Learning (RL) coupled with Deep Learning enabling autonomous control of flows in SDN. Therefore, in this work we propose a routing optimization algorithm based on Quality of Service (QoS) parameters by using Deep Reinforcement Learning for SDN based Data Center Networks. We compare our proposed technique with Dijkstra's algorithm and present our analysis. Our proposed technique achieves 21% better average throughput and 17% lesser average delay compared to Dijkstra's algorithm.