NETS: SMALL: Fault and Performance Management in Carrier-Grade Virtual Networks Over Multiple Clouds

Washington University, Saint Louis MO

Investigators

Abstract

This project seeks to carry out research leading to the development of an intelligent fault, configuration and performance management framework for virtual carrier networks. The research activity will consider improvements to the management framework in a multiple cloud setting so that the high availability and resiliency requirements of carrier grade networks can be realized. To accomplish this objective, the project team will apply novel Machine Learning (ML) algorithms to perform network fault detection and isolation. The research activity has 3 main thrusts: 1) adapting classical machine learning algorithms to target a very dynamic, distributed and real-time environment; 2) developing deductive-predictive fault and performance detection and localization techniques; and 3) developing a hybrid 'central-distributed' technique for training the machine learning models. A key project outcome is to enhance understanding of how Network Function Virtualization (NFV) management, cloud management and operations support can leverage machine learning. The research team will develop new techniques for training machine learning models, and a testbed will be developed to validate the approach.