As someone who played a key role in the development of AT&T's NFV/SDN-based network architecture, Domain 2.0, Tom Anschutz is constantly looking at what's next on the virtualization horizon.
In part two of this Telco Transformation Q&A, Anschutz, distinguished member of technical staff at AT&T Inc. (NYSE: T), talked about the impact of the large web-scale service companies on the telecom sector and how virtualization is evolving with containers and microservices. In part one, he spoke about how AT&T has adopted web-scale practices for its own uses. (See Anschutz on AT&T's Use of Web-Scale IT.)
Telco Transformation: Are the web-scale service companies having a significant influence on the telecom infrastructure, equipment and service sectors?
Tom Anschutz: They have, so I would say "yes" overall. A lot of what they've done, in terms of microservices, is focusing on a sort of computer service model, rather than a functional entity model, which is really core to the whole transition into microservices as VNFs. That's sort of the big thing. They've [web-scale companies] also improved operational efficiency and what I would call maybe "business efficiency" as well.
But in other cases, we don't have as much of an alignment with the large web services companies. Notably, their business typology, or their infrastructure typology, is significantly different. They remain on the bleeding edge of buying more servers, and putting them into a small number of very large locations. They keep taking the technology in one direction, one that we're not close to needing or being at.
The fact that they have these very large locations, they know exactly how many servers they're going to buy, and whatnot, and that leads to a very different business relationship with manufacturers than telcos would typically be able to enjoy. For example, when Facebook creates yet another server -- even when they freely open it and provide it at Open Compute Project -- it's not instantly clear how anyone could make use of it. Because it's a given that it has Facebook's operations, it has their model for acquiring servers in certain numbers. Sometimes it's difficult, then, to see how you make use of that, even when it's been offered up as something to be shared. So it falls on both sides.
There's a good benefit in reuse and learning from what's gone on in the web-scale organizations. But not everything maps up. It's not the same business. It's not doing telco just like web-scale companies do IT. So it does require a mapping, or a thinking through, on how to gain the benefits from what they have invented and developed.
TT: Are web-scale IT practices playing roles in AT&T's CORD, ONAP and Indigo initiatives? (See Pacewicz: Indigo the Next Innovation Wave for AT&T.)
TA: Oh, yeah, absolutely. In CORD, many of our original architecture VNFs were actually microservices. The work that we're continuing with CORD is sort of expanding on that. We're looking to take some of the things, like OpenStack , that were previously less microservice-oriented, and migrate them into containers instead of VNFs, and allow for them to scale down into these smaller edge applications. And similarly -- I guess I went off on a tangent there -- we're working that within CORD. We're also working to obviously add OpenStack. These things become core components that get integrated into ONAP, or what we call ECOMP internally.
TT: Any final thoughts on how AT&T is moving forward with its SDN and NFV virtualization efforts, which includes having 75% of the network software-based by 2020? (See Donovan: AT&T Closes In on Tipping Point With Virtualization.)
TA: One of our themes this year is that we're marching forward fairly quickly.
I'm happy with how we're going with this transformation from legacy elements into a microservices, software-based world in the future. We've taken several steps. The first one I claim, and you've heard this 100 times, is a simple port from the boxes that came before into the X-86 boxes. The basic virtualization. Then there is the second step, where you start realizing that you want to disaggregate those functions into smaller components that are more microservices-based and are more reusable.
I think we're sort of in that transition, and we're looking forward to what comes next, which I think is having access to the source, whether it's hardware specs, or software open source, to be able to modify and address operational and business complexity. I think that how SDN and NFV ultimately address opex is sort of the million-dollar question. We don't have all the answers yet, but I see us clearly marching toward addressing that issue. I'm really pleased with the way things are unfolding.
— Mike Robuck, Editor, Telco Transformation