The Software Defined Networking (SDN) movement has a few core leaders that have been around since it sprouted from the grass-roots academic environments of Stanford University and Berkeley, where we imagine networking geeks called for revolution in nearby coffee houses. Dan Pitt, Executive Director of the Open Networking Foundation (ONF), is one of those people.
Though not exactly a pure revolution — after all, SDN and open networking have already created billion-dollar startups such as Nicira — there is still a core charged with defending its openness from the assault of darker, propietary forces. That’s where Pitt and the ONF come in, guiding standards development and integration.
Pitt spent twenty years developing networking architecture, technology, standards, and products at IBM Networking Systems in North Carolina, IBM Research Zurich in Switzerland, Hewlett-Packard Labs in Palo Alto, Calif; and Bay Networks in Santa Clara, Calif., where he was vice president of the Bay Architecture Lab. Pitt became vice president of Nortel’s Enterprise Solutions Technology Center, spanning nine cities on four continents.
From 2002–2007 Pitt served as dean of the school of engineering at Santa Clara University and holder of the Sobrato Chair in Engineering. Dan received a B.S. in mathematics (magna cum laude) from Duke University and an M.S. and Ph.D. in computer science from the University of Illinois. He has fifty publications and one patent to his credit.
On to the ONF. Pitt joined the ONF March 2011, upon its launch. This was before SDN was even a glimmer in a Cisco marketing executive’s eye. Things have grown more competitive now. Since then, VMware has bought Nicira, networking giants such as Juniper and Cisco have bought their own startups, and Cisco has made SDN a corporate marketing priority.
Does this mean SDN has gone mainstream? When I asked Pitt this question, he pointed out that on a recent transatlantic flight, he read a reference to SDN in an in-flight magazine. He also said that while recently listening to National Public Radio (NPR), host Garrison Keillor made a joke about SDN, playing off its esoteric-sounding name.
Read on, for the interview with Pitt:
Rayno Report: Okay. So, is SDN that big a deal?
It is a big deal. It will take a while to become the modus operandi of the network. It’s definitely headed in that direction. Nobody’s going to put the toothpaste back in the tube.
The business leaders understand what’s at stake, they’ve got to do it to be successful. I was on a panel last year with CTOs of major service provers. One of the them understood the whole thing, the other one was skeptical. Then he turned to me and asked how he would be compelled [to implement SDN]. I said the competitors will compel him, they will offer services faster. It’s going to change the way a lot of the operations staff operate their companies. The processes and technologies will change, and the skills will change. Any time there is change there is disruptive.
More and more of their operations will be data centers. Their network will be more basic switching equipment. With SDNs, protocols go away. Essentially the infrastructure becomes software. The software runs in the software data center.
RR: Let’s talk about some of the origins of ONF. How did it start?
Pitt: We were started with network operations at Stanford and Berkeley. When OpenFlow protocol was [invented], they said we needed to standardize this and get it launched, but we needed to become the masters of it so that the vendors couldn’t tell us what to do.
So they created ONF. When we launched in 2011 I became a full time. We have over 2,000 participating in our market organization. We’ve been asked by a lot of people to help them with SDN. That’s why we work with OpenStack, Open Data Center Alliance, OIF, and ETSI/NFV.
NFV is something the operators can understand because they buy as many appliances as switches. They’re all specialized, single function, and expensive. We thought this was such a great use case for SDN, so we signed a memorandum to help make them successful.
If you are doing load balancing, you need to [connect with] the control plane. SDN is the connection between that. We are helping them. They have a big interest in service chaining. We’re helping them with that.
Our goal is successful commercialization of SDN. In the last 3-4 years it’s made amazing progress. We know it’s a long road, we celebrate the milestones to keep everybody enthused.
RR: Tell us more about how this should effect service providers.
Pitt: The carriers have two avenues of exploiting this. One is for their own data centers, the other is to offer cloud services. They sat out cloud services.
We will see a stratification of operators. Some operators are good at customer interaction and billing and incentives, others enterprise services, others are good bandwidth optimization. Many of them are virtually integrated, but none of them are [focused]. They will end up being focused on what they are very good at.
RR: Where do you see the networks in 5-10 years?
Pitt: There is a big difference between five and 10. We will see integrated control of packet optical and wireless. Operators will be optimize for Ethernet and IP. It will be much more efficient, there will be much more choice.
The infrastructure will become much more basic, there will be packet forwarding at the optical network. They will be more flexible in offering services. We will be able to embed storage in processors. Data centers large and small will be confederated throughout the network and the data center becomes a collection of compute, storage and onnectivity that can be composed to conduct different functions, they’ll find that they will have lots of room to innovate.
It’s going to be a race because we have the over the top (OTT) providers, Facebook, Googles, Amazons rolling stuff out not worrying about long-distance carrier stuff. That whole side is looked at askance by the service providers. But you can mock Google and Amazon at your peril.
RR: Yes, the critique of service providers is they tend to be more bureaucratic.
Pitt: They won’t survive with that level of bureaucracy.
RR: What is some of the coolest stuff you’ve seen lately?
Pitt: Somewhere on the other side of the Pacific, I won’t reveal the name. An Internet Exchange is built with several peering points, using BGP between them. Without changing the gear they have they overlaid with OpenFlow allowing them to prioritize their traffic by application. It was easy and fast and non-desruptive.
RR: That sounds a little bit like flow-based routing, something we’ve heard about through the years.
Pitt: Flow-based routing is not even possible with IP. It is based on IP. It’s a whole different way of the networking. [OpenFlow] gives you more granularity of flows.
This is old news, but Google’s G4 network, it connects all their data centers. It went public two years ago and they raised WAN utilization from 20% to 90%.
There will be new ways of handling entertainment video and caching, in a more efficient and inexpensive way than Akamai. There is a service provider doing this. Looking at the scale that some of these web service operators are building and some of the mobile operators are building, they couldn’t do it this fast any other way.
RR: Thank you Dan. [Fade to NPR theme song.]