Thomas D. Nadeau and Ken Gray are both distinguished engineers in the Platform Systems Division (PSD) CTO Office at Juniper Networks. Tom Nadeau is responsible for leading software-defined networks (SDN) and network programmability. He also is an adjunct professor of computer science at the University of Massachusetts at Lowell and is on the technical committees of several prominent networking conferences. Ken Gray is responsible for technical strategy and innovation for Juniper’s PSD, with a focus on core routing and software-defined (driven) networks.
SDNCentral has made arrangements with O’Reilly Media to provide an excerpt of Nadeau and Gray’s software-defined networking book, SDN: Software-Defined Networks – An Authoritative Review of Network Programmability Technologies, to our members. Or, buy SDN: Software Defined Networks at Amazon today.
Gray: Tom and I both have been working on the OpenDaylight Project (ODP). It overlaps a lot with the work we do at Juniper, where for the past year we have been giving talks on SDN to customers, trade shows, and standards bodies. We’re also getting involved in various aspects of standardization in the IETF (Internet Engineering Task Force). For example, we’re collaborating and contributing to the work in I2RS (interface to the routing system). We’re also involved in the drafts on network service chaining, which is related to the ETSI (European Telecommunications Standards Institute) NFV (network functions virtualization) interest group.
What was your main goal in writing the SDN book?
Nadeau: The goal of the book was to try to level what’s going on in the industry. There are a lot of moving parts around SDN, and there seem to be more every day. As I began work on the project last August, I realized there was a lot to write, and I roped Ken in.
Gray: We wanted to talk about where we’ve been with SDN, what’s going on in the marketplace now, and where we think it’s going. We wanted to get to the basics of what SDN is about. We agreed SDN ultimately is about the programming environment, and ancillary to that, about management.
Who is the target audience?
Nadeau: We’ve talked to a lot of people in the last year who were tasked with figuring out how SDN is going to save their organization money and how to get involved in it. Most of them are lost on where to start and where things are going. The book is really targeted at them. We want to help people separate unicorn and fairy dust from reality.
Gray: Part of our target audience is folks trying to understand the lay of the land, so we tried to make the book not too excruciatingly technical. We gave an overview and went down one more level with references to dig deeper.
What was the hardest part in writing the book?
Gray: It was hard knowing when to call it quits. We were just starting with ODP when we began writing the book, and NFV was heating up. One of the hardest things was knowing there was even more information to come and saying, we just have to stick a stake in the ground and go from here.
Nadeau: It was also tricky to try to write with one voice, since there were two of us, but we think we managed quite well.
Which chapter is your favorite?
Nadeau: For me, it’s between chapters four (“SDN Controllers”) and nine (“Building an SDN Framework”). It was good to remember the totality of the space and why each of the controllers was important not only in its own right, but also in how it pushed the envelope in the development of others afterwards. In much the same way, we talk about the OpenDaylight controller as a culmination of the avant garde, or the latest incarnation of that progression of controllers.
Gray: For me it was chapter nine. I really liked writing about the experience we had in building a framework and this revelation that it didn’t have to be a single southbound protocol. We started realizing you could do more if you got around the religious war of which one was best and recognizing that they all have a particular purpose.
What’s your take on the market evolution so far?
Gray: The infrastructure that manages the state has been an evolving thing. The answer isn’t who makes the best controller – it’s who made the most recent one. We all learn from each other, and nobody wants to admit it, but in this environment you always steal the best ideas that are available out there. The most recently developed controller model has a modular architecture.
The application programmer and the consumer shouldn’t have to care about whether the abstraction is hop-by-hop OpenFlow programming or a VXLAN overlay. The control protocols all have different strengths. We shouldn’t eliminate them by rote right from the start, but rather, let them develop.
We’re still in for a bit of a hustle with distracting arguments over what the best controller architecture might be. I think as an industry we eventually will come to the conclusion that multi-vendor supported open source is the way to go. Then we can take off from there and start focusing on application portability.
What’s been the most surprising thing you’ve seen as vendors and markets evolve? The most disappointing?
Nadeau: The most surprising thing I’ve seen is the rapid evolution of SDN. When I started working on SDN two and a half years ago, it wasn’t even really called SDN. To go from where we were at that point to where we are today is pretty amazing.
I worked on MPLS (multiprotocol label switching) when it was in its embryonic state, and it took quite a bit longer for it to evolve. We came at SDN from a different perspective. SDN is not something that’s driven by the carriers. Instead, it was the academic and large web provider guys who were actually using these concepts. They just went off and solved the problem – necessity being the mother of invention.
Gray: For me the most surprising thing was the open source movement and how people realized we’re not going to get anywhere if we all make separate controllers. At least I hope that’s what they realized.
As for disappointment, if we’re trying to solve the problem that servers have had with their multi-OSS/BSSs (operational support system/business support system), we don’t solve it by creating a whole bunch of vertical stacks of applications, bound to controllers, bound to a particular protocol that talks to a switch. That doesn’t move the ball. And yet as soon as SDN became popular, the reaction was the proliferation of a whole bunch of these things. We have vendors coming out with essentially proprietary solutions saying, “Hey, come join my ecosystem.”
Well, you can’t have ten different ecosystems. It doesn’t fit most software developer’s business models. Most people who are working in the enterprise space and consulting or writing software can’t survive in that space.
Do you think OpenDaylight Project will dominate the controller marketplace?
Gray: If you look at the number of companies that are involved and the number of people committing code, the ideas coming out of ODP are really quite impressive. The question is whether you can fit all those ideas into a first release.
It puts things in their proper place to make the controller open source because it really goes to the value – even in the highly integrated vendor offerings – that’s in the applications.
What can we come to expect in the next 12-18 months in the SDN and NFV ecosystem? What’s your boldest prediction?
Gray: I think obviously OpenDaylight is going to ship product. It’ll be a first product, and you’ll see a lot of bantering back and forth over whether it’s perfect or not. A lot of people don’t realize that it’s not until after we ship that a lot of the community actually kicks in. You have to set your expectations appropriately.
I think we’re in for a year of distraction. The ETSI NFV folks are getting to the point where they’re going to publish their recommendations, but already I see they’re kind of drifting away from their charter, like working on the VNFD (Virtual Network Function Descriptor) that doesn’t exist today and should be part of a new protocol.
Meanwhile, there are those that will still come out with proprietary solutions. Service chaining is the new hottest thing, so vendors are going to do it in a proprietary fashion, using whatever protocol is their favorite protocol.
Until everything is open, you don’t want to buy proprietary solutions unless the ROI is significant.
Nadeau: If you talk to me a year from now, I would hope OpenDaylight is shipping not one, but several releases and that there would actually be enterprise versions of those releases being supported by equipment vendors and software vendors.
The other thing I think will be happening is we will start standardizing the APIs and functions and use the hard work that’s gone on in OpenDaylight to drive that. Finally, I think we will start to see people talking about full-blown SDN deployments in large enterprises. I know many large customers that have been experimenting and toying with some things. I think we’ll start to see them announced and used and getting into the mainstream rather than being a kind of exotic sideshow.
Thank you, Tom and Ken! Wish you the best of with your new book.
Tom and Ken’s book is available for purchase on Amazon:
The comments expressed by the interviewees are their own and do not reflect the view of their employer, or O’Reilly Media, the publisher of their book.
Check out more Interviews on SDNCentral: