January 22, 2020

239 words 2 mins read

How to use Machine Learning for Testing and Implementing Optical Networks

How to use Machine Learning for Testing and Implementing Optical Networks

Optical transport networks are evolving to have unprecedented flexibility with advances such as finely tunable bitrate transponders, allowing dynamically-adaptive …

Talk Title How to use Machine Learning for Testing and Implementing Optical Networks
Speakers Stevan E Plote, Nokia, Jesse Simsarian, Nokia Bell Labs, Peter Winzer
Conference NANOG71
Conf Tag
Location San Jose, CA
Date Oct 2 2017 - Oct 4 2017
URL Talk Page
Slides Talk Slides
Video Talk Video

Optical transport networks are evolving to have unprecedented flexibility with advances such as finely tunable bitrate transponders, allowing dynamically-adaptive operation near the fiber Shannon capacity limit at near zero system margin. How Network Operators and Content Providers can take advantage of this without having a staff of optical experts to drive network performance optimization is by using machine learning. Furthermore, the development of network operating systems (OS) enables network programmability and support for multi-vendor network elements. Network operating systems lay the foundations for advanced machine learning algorithms that operate on an abstracted network representation presented by the network OS. Refinement of the network model parameters improves the machine learning results and its representation of the actual network. In this talk we will discuss how network sensing, machine learning, and actions taken by the network OS can lead to a more optimized network that can efficiently support traditional and cloud network services. We will also point to challenges that will have to be overcome to make such networks commercially viable.

comments powered by Disqus