Machine Learning Driven Model for Software Management of Photonics Switching Systems

Modern elastic optical networking requires additional flexibility at each layer compared to the traditional approach. The application of the Software-defined Networking (SDN) paradigm can provide the required degrees of freedom. The implementation of optical SDN down to the physical layer requires the complete abstraction of network elements to support full control by the centralized controller. In this work, we propose a topological and technological agnostic model based on Machine Learning (ML) to abstract the behavior of optical switches for the computation of Quality-of-transmission (QoT) penalties and the definition of control states. Training and testing datasets are obtained synthetically by software simulation of the photonic switching structure. Results show the capability of the proposed method to predict QoT impairments with high accuracy, and we envision its application in a real-time control plane.