Collaboration and common approach needed for optical network standards

Adrian Farrel, Managing Director, Old Dog Consulting, delivered a presentation on how Machine Learning (ML) can improve the performance, reliability, and scalability of optical networks. Farrel touched on the presentations before his and advised that if we do not have access to all of the data then no amount of ML will help enhance fixed networks, and underscored the key roles that vendor-agnostic standards and interworking play.

“We need to put all of the building blocks together before applying ML. I avoid saying ‘Artificial Intelligence’ as it is a ubiquitous term, has too much industry hype, and lacks focus. ML is a better phrase as we are not yet talking about algorithms that self-tune, but sophisticated algorithms that spot patterns and suggest relationships,” Farrel advised.

Farrel highlighted the use cases of performance monitoring and fault detection for optical fixed networks that ML can enable. He outlined the importance of ML acting as the glue for data profiling, rapid response, and taking predictive actions before a problem arises. This has the potential to deliver greater Return on Investment (ROI) through automating network operations, intelligent fault diagnosis, and anomaly detection.

Farrel talked about how the IETF can collaborate with the Broadband Forum to develop universally adopted standards for ML data models. The IETF’s Abstraction and Control of Traffic Engineered Networks (ACTN) framework offers a unified approach for interworking between software and hardware vendors. ACTN brings enhanced scalability and flexibility through dynamic network resource allocation.

ACTN and the Broadband Forum’s CloudCO project have similar objectives, and their openness and programmatically exposes simple and common interfaces, Farrel advised. This would be an ideal opportunity for the Broadband Forum and the IETF to come together and offer a global view, “rather than implement two sets of management models with common components.”

While optical networks have been managed with five considerations in mind: fault, configuration, accounting, performance, and security, the initial ACTN work only focused on provisioning and virtualization. But Farrel pointed out that Fine-Grain Network Management could be integrated into an ACTN architecture, building on existing IETF YANG models and offering migration to a full YANG-based management system.

When it came to IETF YANG modelling, lessons had been learnt along the way. There are IETF YANG models for everything and the ‘cookbook’ is getting large, Farrel added. A simple, generic base model that can be extended and augmented for other technologies, allows maximum reuse for implementations, and provides multiple vendors with a common approach. This would open the door to those with specific expertise, such as the Broadband Forum, to get involved and construct access-specific models based on augmented IETF models.

The IETF’s Event Condition Action (ECA) framework can be integrated with ML to recognize behaviors and data patterns, predict faults, and diagnose the root causes of faults. Farrel advised that the IETF was working on a detailed set of control, configuration, and management YANG models to report and inspect incidents. By integrating ML into the framework, Farrel stated, routine tasks could be automated, network resources dynamically allocated, and predictive maintenance carried out. This could lead, in time, to self-managed networks and more efficient network operations.

Farrel compared self-driving vehicles to our broadband networks. “When it comes to vehicle safety, the car stops. But if our broadband network suspects something strange, we do not want network traffic to stop.

“Network operators did not initially trust software to control our networks and human interference will continue to exist for a long time. But once we get comfortable and have completed a number of field trials, networks will be free to run by themselves. Collaboration and commonality of broadband standards will be key,” Farrel stated.