Disaggregation vs. integration of the OLS

Paul Morkel
Sliced apple

I was in Texas last month at the inaugural Open Optical Conference co-facilitated by PenWell Corporation and Lightwave. There was a particular focus on open optical networking, which gave delegates a chance to discuss the specific characteristics of photonic networks and the impacts of OLS disaggregation. What became clear was that there are very different camps in the open optical networking space with regard to the question of disaggregation of the OLS.

Glenn Wellbrock, director of optical transport network architecture, design and planning at Verizon, gave an account of the topological complexity of their network today. He went on to describe very clearly why they see the optimum solution as being an integrated solution. On the other hand, Andrew Leong, Facebook’s optical network manager, described the company’s plans for disaggregation at the ROADM-to-ROADM or OMS level.   However, one thing most representatives were able to agree on was the need to avoid lowest common denominator standardization for disaggregated optical layer performance. There was also a shared appreciation of some of the unique analog characteristics of the OLS.

YANG model agreements are an important part of enabling a common centralized control model. It’s clear that OpenConfig, which is favored more by the cloud provider community, and OpenROADM, which is driven by AT&T and gaining traction in the carrier community, are the two main device models today. Each of these is evolving to describe more of the required functions of the optical layer including Flexgrid. Agreements on APIs are also a necessary component. In this area, NETCONF/YANG has the most traction to date, although binary interfaces such as gRPC and gRPC network management interface (gNMI), which also supports YANG models, look to be rapidly advancing.

The outlook may then be that we see both approaches being adopted – integrated networks as advocated by Glenn and also disaggregated OLS as advocated by Andrew – depending on the unique characteristics of the network operational environments. What will be needed are OLS solutions that can work in either environment.

Related articles