Lisa Caywood was at the ground floor of software-defined networking (SDN) and she had a pretty good idea of where it was headed years ago.
Caywood, director of ecosystem development for OpenDaylight, spoke about the evolution of SDN during Part III of this Telco Transformation Q&A, as well as the development this year of northbound APIs and policy across the industry.
In the second part of the Q&A, Caywood discussed a recent tour of China, and how open source groups could do a better job of collaborating with each other. In the first Q&A, Caywood talked about last year's Boron release and OpenDaylight's plans for this year. (See ODL's Caywood: Why China Matters and OpenDaylight's Caywood Discusses 2017 Roadmap.)
Telco Transformation: How will SDN evolve going forward?
Lisa Caywood: I think that I wrote about this three years ago actually. It's really funny because you can go through all of my old blogs
and in most cases I would say the exact same thing today or I would say "Look I was right." I said that back then controllers would become one of two things. Either they will migrate up and become part of the management and orchestration layer, or they will be subsumed within the network operating systems of the various equipment vendors, which will now basically have better APIs and more sophisticated programmability than before, and that's exactly what has happened.
I think that there's clearly an increasingly strong emphasis on the orchestration layer and that totally makes sense. In fact to go back to last year's Boron release. If you look at the Boron architecture graphic, most of the new projects were kind of in the specialized features bucket. There were a few in the standard control plane functions and very little in the Model-driven Service Abstraction (MD SAL) Layer. Not because there wasn't work being done there, but because the work that needs to be done there is very incremental or it's more about insuring the robustness of the platform, the stability, the security and things like that as opposed to adding new features.
So over time more and more of the new projects and more and more of the developer focus is going to be on the higher and higher levels within the platform because now the foundational levels are relatively mature and stable. And what is less mature in the networking industry as a whole is how do we -- and this kind of goes back to the core of why SDN was thought of in the first place -- effectively manage a heterogeneous network environment? Being able to gain an end-to-end view of that environment and make intelligent preemptive, proactive decisions about the network traffic based on that end-to-end view.
That's why I think that now that we have a foundation in place the focus is going on increasing north bound interactions with the management plane and the orchestration layers. We've seen that with a lot of work with OPNFV and OpenStack, with OPEN-O becoming increasingly mature and with ECOMP being opened up.
It's not that the orchestration stack makes it moot. The higher level orchestration frameworks don't do the deep level network tooling and management that controllers do. Controllers are an integral part of those larger orchestration stacks and frameworks. They are the ones that are headed down into the bowels making sure that all of it actually happens. The higher-level frameworks just don't have the deep tooling, from a network perspective, that the controller layer does. But as they say, it's something that gets subsumed either in the network OS or at the key component of the orchestration layer.
Want to know more about the companies, people and organizations driving developments in the virtualization sector? Check out Virtuapedia, the most comprehensive online resource covering the virtualization industry.
Will the industry see more clarity on northbound APIs and policy this year?
LC: That's sort of in the category of ongoing work. The new (ODL) projects depend very heavily on those things. In fact, one of our lead contributors on intent, Yrineu Rodrigues, published a blog on this recently. I expect we'll be seeing a lot more there as it relates to the other projects we're doing like analytics, predictive learning and things like that.
TT: Will the "controller wars" continue this year, or will orchestration stacks make them moot?
LC: Aren't we done? It felt like the controller wars ended a year or so ago. I think at this point things are shaking out enough that it's fairly clear who is going to choose what for why. There's probably a little more clarity that can be driven, but some controllers, especially the ones coming from the proprietary space, tend to be network management ++ in that they're sort of advanced element managers. They're built to run the equipment that the vendor also sells. On one side there's the (VMware) NSX model that is geared towards a very specific set of use cases, but also much more towards the server admin as opposed to the network admin.
Somewhere in between there is the open source world. Telcos have jumped on that by going whole hog. I think more traditional enterprises are more accustomed to having things that they can just take out of a box, stick in a rack and plug in. So open source is something that is a little less comfortable for them. I think as the platform has matured the user interface has improved, and the commercial offering has matured so now its becoming much more consumable. In fact, we're seeing a number of enterprise customers joining us now. When I say enterprise that means everything from high-end financial institutions to state and local government.
If anything, the state and local healthcare, or anything related to IoT in particular -- where the needs of the network aren't traditional and they're not defined within traditional network management tooling -- those seem to be particularly attracted to open source models because they can pick and choose their own features and their own capability. If there's not something you need, you can just go build it.
So there have actually been two or three of the startup members we have focused on that space. They have a very clear opportunity around selling customized solutions and value added solutions for those new emerging types of use cases that don't have the traditional tools built into them already.
— Mike Robuck, Editor, Telco Transformation