Soft Failure Detection, Categorization and Localization

Failures in the optical layer can affect the operation of the network as a whole, and the services it supports. The current trend in optical networks is to reduce the operating margins. This means that even relatively slight degradations of the Quality of Transmission (QoT) of the lightpaths can affect the quality of service of the whole network. In this paper we thoroughly evaluate various Machine Learning (ML) algorithms to detect, categorize and localize failures. We leverage information from different QoT metrics to improve the accuracy of the algorithms.