The Open Networking Foundation (ONF) is the public face of the Software Defined Networking movement, spelling out requirements and defining standards. The group's board includes Google, Facebook, Yahoo, Microsoft and Goldman Sachs on the data center side, and Verizon, Deutsche Telekom and NTT Communications on the service provider side.
Additionally, there are close to 150 members, from global telcos to startups. To get a sense of where the movement stands, John DIx, Editor in Chief of ComputerworldUK's sister title Network World, tracked down ONF Executive Director Dan Pitt, who spent 20 years developing network architecture, technology, standards, and products at IBM Networking Systems, Hewlett Packard and Bay Networks.
How would you characterise what the ONF has accomplished to date?
ONF has successfully championed the notion of SDN that everybody is lining up behind, regardless of their definition of it. Everyone is saluting the SDN flag. In addition, we have championed and standardised aspects of and built-out architectures for the OpenFlow protocol, which is the sole vendor-neutral way to connect the forwarding plane and the control plane.
We have also taken many steps to advance the experience of adopting and deploying SDN with our plugfests, our conformance testing, laboratory sanctioning and through the work we're doing with everyone from chip vendors to software solution architects. We feel a responsibility to drive the whole movement forward, and I think we've succeeded, with vendors wanting to abide by the same principles we have been advocating, of openness and multivendor inclusion and a real focus on software as being the center of value in networking.
Are you where you thought you would be by this point?
I didn't have explicit expectations of where we would be in three years. We had expectations of the first year and maybe into the second year. We exceeded those by every measure, including the scope of our technical work activities, the scope of our market education activities, and just the sheer numbers of both member companies and individual participants. And even in the mindshare that seems to have travelled around the world. I was at a conference in Silicon Valley and saw a terrific presentation on SDN security by a researcher from Northern Ireland, and she was totally conversant about SDN, the principles and what we're trying to do and where security comes into play. We've got these great minds all over the place working on these issues and problems, and advancing networking in this way. This is nothing we could ever imagine doing alone or ever expected to. I'm astonished every day at the caliber of the brain power that's being applied all over the place.
Anything that you didn't achieve that you had hoped to achieve by this point?
Good question. Our success measure as an organisation and my personal goal is to see SDN implemented, adopted and deployed. That's happening in lots of corners, but it's taking longer than I wanted. But it's not longer than I expected. What I've come to understand is that it is way more than just a technical change. We knew it was a new way of thinking about networking and computing combined, but it is also an organisational and a human challenge, because the roles of IT organisations and the people therein are changing. We're very interested in the training aspect, the skills aspect, and helping organisations get there from here.
Version 1.4 of OpenFlow is out now, and some analysts call it the first version that's ready for prime time, but would you agree with that statement?
I think 1.3 was ready for prime time. That's why we declared it a stable target for implementers anywhere. Software people are implementing it and deploying it. But 1.4 continues to improve on it and we're working on 1.5 beyond that. Prime time very much depends on your application. We have people that are thrilled with OpenFlow 1.0. I think the diversity of applications and requirements, and the diversity of value for the different kinds of businesses that will benefit from SDN, is so wide it's hard to say "this is the ultimate answer for everybody."
When does 1.5 come out?
We're looking at about one a year. And every time we do a new release we are taking the new features and bundling them as an extension package to OpenFlow 1.3, so people who made use of that as a stable target will find it easy to adopt.
Next section: WHy some vendors are spurning OpenFlow
I presumed by this point, three years in, that more of the vendor community would be embracing OpenFlow. Some have thumbed their nose at it, saying they'll add support if and when needed. What do you make of that?
There are two reasons why you might see this response from the vendors. One is the hardware vendors had some legitimate reason to say it's been hard to implement OpenFlow beyond 1.0. But we addressed that with the later versions. And the other is that, when the playing field has been leveled, you have to compete on price, quality and features. It creates a more competitive market, which is great for customers, but it does change the business models for some vendors.
We understand that this is a transition for them. We're trying to keep it going as fast as possible. It's like we're pulling this gang with an rubber band. We don't want it to be a rigid pole because they can't move as fast as we move, and we don't want it to be stretched so far that it breaks and we go on and they don't have any incentive to follow us. So when it's something elastic we can surge ahead and they will catch up. More or less we do move forward together. It's not instantaneously at the same rate.
Good analogy. How about with Cisco? They want to leave some control in the network gear, so they are not really embracing the full decoupling idea.
That's certainly not the philosophy we're advocating. We believe that having this logically centralised control will bring all three of the benefits that operators want, CapEx savings, OpEx savings and increased revenue velocity. But they are making business decisions-- every company is -- and finding their way to get there from here. Cisco has been a member of ONF since the very beginning and they have participated very well. But they have their own business decisions to make and their own path to follow.
What do you make of Cisco OpFlex, which it describes as a policy framework "to transfer abstract policy from a network policy controller to a set of smart devices capable of rendering abstract policy"?
Our philosophy is that we want the network elements to be as simple as possible so they are just high-performance packet-processing devices. They need to be able to interpret models. We don't think they need to have a lot of custom ASICs in them, and we see the value of being in software up on the control plane, and so we are pursuing all the means we have to make the philosophy that we hold available in products and services.
Then you have VMware, which is pushing network virtualisation, which again is a bit different than your SDN vision. Do you view network virtualisation as an application of SDN, as some say?
I think it's a use case of SDN. It comes in handy in lots of scenarios and has been very useful for a number of organisations to get started. But some users say, "Virtualisation is nice, but I'm not doing overlays. I'm going to do real SDN. I'm going to control the whole forwarding plane."
And as we build out the OpenFlow substrate, we're getting the conversation around to what is truly important to the enterprise, and that's the stuff you build on top of it. Some people debate whether this protocol is better than OpenFlow for this or that use, and I say, let's get beyond that. The outbound protocol is just a vehicle, just a carrier of these things. The real importance to enterprises is the orchestration, the automation and the central policy management. This just carries those commands down to do those things. It's figuring out how those couple to organisational objectives that really matters to people who operate networks.
So we have quite a number of activities in that area, and we're going to see, I think, great progress in helping organisations put these pieces together so they have a functioning system that does bring them the business value that SDN promises.
We continue to work with VMware. They have championed and deployed the overlay approach, which a lot of people have liked and it's gotten SDN in the door. We think there will be some significant industry battlegrounds for giving customers what they want and, as always, the market decides who wins.
Next section: Is OpenDaylight critical to SDN in general?
Switching gears a little ... how critical is OpenDaylight -- the vendor-lead effort to build an open source SDN controller -- to the success of the SDN movement?
We certainly believe that as networking becomes more a software play, open-source software will have an increasingly significant role. So we cooperated in a number of ways with OpenDaylight, and OpenDaylight has a lot of control modules that are useful and important and we're delighted to see people contributing to those.
They have a lot of southbound protocols. We're not sure the industry needs to spend that much time on those. It's better if they focus on the orchestration, automation and central policy management. So we're making sure their OpenFlow southbound protocol option is of high quality and available.
They're targeting particular use cases, which is exactly what they should be doing, but not all use cases. We're cooperating particularly with them in the architecture for the northbound interfaces.Given most organisations have loads of money invested in legacy gear, what are the incremental steps they can take to get there?That's a great question. That's why we started our migration working group about a year and a half ago (see "A Q&A with the Chair of the Open Networking Foundation's Migration Group"). And we are showing examples of how people have done it. We are building some tools and some metrics. Basically there is a combination of things. There's a way of introducing SDN software with existing infrastructure. You don't get all the benefits of SDN by any measure, but you get some, and you learn about how to couple your business priorities to some of the orchestration software.
You gradually move into a forwarding plane that is optimised for packet processing and does not have the legacy functions. And you can introduce this a network element at a time. The vendors are going to choose a variety of ways of doing this with either stand-alone devices or hybrid devices, and there will be a lot of experimentation to do that. But what you want to make sure you have is some understanding of what's happening in your network. That's why we work on configuration of management methodologies integrated into existing management methodologies.
So it's really a combination of deploying new software, coupling new management schemes and deploying new switching hardware, and then it's going to be understanding that more and more servers and switches are built on exactly the same components of Ethernet, x86 and OpenFlow.
I see that in October in Germany the ONF is going to be showcasing some SDN solutions. Will those be actual users up there speaking?
The SDN Solution Showcase will be part of the Expo area and we're putting together six or seven themes and I'm not ready to publicise those yet. But every theme we demonstrate will involve at least two vendors and one [user]. So this is different from the normal Expo where every vendor goes out by himself and shows his own stuff.
Anything else that I didn't hit on that you think the world should know about ONF and OpenFlow and where we are?
One thing that hasn't changed in the three years that we've been around is our mission to accelerate the adoption of open SDN for the benefit of users. And I can say that we're really grateful for the efforts of everyone in the industry in contributing to making this a successful movement. It's been amazingly successful so far. It has a long way to go, but the benefits are so clear to people who operate networks that there is just no turning back.
I think now that people are starting to understand that networking is very much about software, we're going to see more emphasis on software skills in arenas that used to be just about networking. So we're very happy and inspired to be leading this movement. We take our obligations very seriously. We're working hard to make it successful for operators and users everywhere.