Dynamic controller deployment in SDN networks using ML approach

Abstract

The Software Defined Networks (SDN) architecture deploys the programmable network by decoupling the data plane and control plane from the existing network architectures. Control activities are put into a software called controller. This new architecture, utilizes programmable controllers, enhances the intelligence of the networks’ operations and enables network engineers to serve their business requirements more efficiently. One of issues in SDN is, estimating the required number of controllers needed and placing it in optimal locations. Many works have been proposed to place controllers in its optimal locations. In most of the works, the controller placement was based on some mathematical formulations, or by heuristic approach and number of controller required was given as an input parameter. In this work, a Traffic Engineering (TE) based controller deployment is proposed. For placing controllers K-Medoid algorithm was used and ANN model was created for analysing and predicting the traffic.