Thursday, 07 March
12:30 - 14:00
Expo Theater I
Moderator: Tad Hofmeister, Network Architect, Google, USA
New network analytics frameworks, extensively based on new streaming telemetry methodologies, and more recently combined often with Machine Learning (or even Artificial Intelligence) for the potential of cognitive systems, have been increasingly considered an important evolution of network management and control for wire line transport. More specifically, for example, network operators have developed a few new related wireline transport automation and abstraction frameworks, like Openconfig, Open‐ROADM, or TIP that have enjoyed significant interest, and the first two are already being extensively adopted.
The underlying motivation for this evolution in network management and control is to a great extent aiming to replicate the automation use‐cases successfully employed by hyper‐scale compute, predominantly in the data-centers of the public cloud infrastructure. These innovations have been extended, the last four to five years, to networking with the exciting end‐goal of a fully autonomic, policy‐driven operations paradigm with little (if any) human intervention. However, extending the automation achievements of compute to wireline transport pose a few interesting new challenges. Notably, optical transport networks are characterized by significant heterogeneity in technology, failure modes, and performance metrics; e.g. latency variation can be three to nine orders of magnitude more than in compute while availability requirements are much stricter (typically five 9s).
This panel will debate the value, current reality, and future promise of streaming telemetry, analytics, and cognitive systems in transport network management and control. Among the many interesting related topics, the panel will particularly aim to debate:
- What are the key enabling innovations, and remaining limitations towards this new generation of network management for wireline transport based on Streaming Telemetry and Network Analytics?
- What is the current reality, and realistic future potential of Machine Learning, and Cognitive Systems in Transport Network Management?
- What are the key similarities and differences in network analytics and cognitive systems between packet and optical transport?
Tad Hofmeister, Network Architect, Google, USA
Tad Hofmeister is a Network Architect at Google with a focus on scalable, cost-effective DWDM technologies and software management innovations. He is also an active participant and board member of the Optical Internetworking Forum (OIF). Prior to working at Google, Tad was an architect, system engineer, and hardware designer for several optical transport and packet processing companies including Ciena, Matisse Networks, OpVista, and Applied Signal Technology. Dr. Hofmeister earned his Master of Science and his doctorate degrees in electrical engineering from Stanford University and a Bachelor of Science from Columbia University and Bates College.