Dynamic Routing Framework Proposal for SDWAN using Topology-based Multitask Learning

Networking is a necessity in today's world of connections and the need for efficient network routing is driving the use of Software Defined Network in a variety of corporations for their high performance and security. The architecture of Software Defined Wide Area Networks is a high-level structure of software and hardware scalability for deployment. In this paper, a SDWAN network with Clos topology is implemented and the dynamic routing data is collected from this network. The derived flow tables are analyzed and a Deep Neural Routing methodology is proposed for the optimization of data flow via route predictions. By predicting the route for data packet routing over the network for multi-level switches using Multitask learning higher efficiency with distinctive security requirements can be achieved. The network architecture constructed spans over multiple clouds whose boundaries are set by the cloud deployment model.