Network Routing in SDNs Using Topology Based Multitask Neural Modelling

Software Defined Networks (SDNs) are adapted for their high programmability and security offered by them for use in enterprise networks. The SDN Controller present in these networks have the function to route the packets in the network to enable communication between the end-point devices connected to the network. In this paper, we created an SDN environment in Mininet using Ryu Controller to collect the communication data to build a topology-based system that could perform dynamic deep neural routing in the network. The neural routing model implemented here is a multitask learning model hosted on top of the SDN Controller that will perform network routing. The trained neural network had given a high-performance accuracy of 99.73% and the proposed system is compared and measured with the existing network routing algorithm in SDNs.