More than steak and salmon were on the menu Tuesday when the SDNCentral team joined Arista Networks and Apollo Information Systems for an educational, pragmatic lunch seminar on SDN. The order of the day: covering SDN fundamentals and taking an honest look at how SDN is working (or not) in the real world today.
Arista Vice President of Marketing Doug Gourlay opened the seminar with a quick look at what is SDN. While purists see SDN strictly as a separation of the control and data planes, and pragmatists associate SDN with programmability, the common view of SDN has come to equate it with network virtualization. (Roy and I would add that SDN equals “services-defined networking,” but more on that here.) Given where SDN is at today, Arista has come to see SDN as a software-to-infrastructure interface that allows applications to drive infrastructure actions.
The excitement surrounding the possibilities of SDN and virtualization has created a buzz not seen in the networking space in some time. But Gourlay reminded attendees that centralization of the control plane for its own sake is not enough to solve actual business problems. IT administrators need to be careful and practical about how to apply SDN in their own networks, tying any SDN efforts directly to solving a specific business issue. Gourlay cited the Open Flow scalable network TAP as one current SDN application that generates ROI and saves money.
“The SDN challenge today,” Gourlay said, “is not to figure out how SDN will replace existing networks, but rather to create a heterogeneous model of SDN that takes existing infrastructures and uses SDN to program exception-based policies on top of them.” Most hardware switches today are not designed for a pure, all Open Flow environment anyway. What SDN can do is slim down operating systems, reducing the clutter on the edge by centralizing control and how topology happens.
After Gourlay’s informative presentation, we took over to look at where the SDN market is today. While SDN in visualization and TAP are showing early promise, SDN is not yet ready for production for most other use cases. We liken SDN today to the early stages of the enterprise software market. Right now, almost any SDN solution is a custom solution. We expect to see that move more toward tailored solutions in the next 12 to 18 months, and ultimately packaged software further down the road.
As for the immediate future, if you are interested in SDN for your organization, learn about it, read up on it. If you have extra money in your budget, buy an SDN switch or controller and play with them. Like Gourlay, we also emphasize the need to establish valid business use cases where SDN can bring value – look for low-hanging fruit to start on SDN. Finally, work with your trusted vendors and advisors to design a POC. We’re seeing more openness in the industry on both the vendor and customer sides than we have in years. Use the learnings you get today to determine appropriate next steps in 2013, early 2014, when we expect SDN vendor products to mature.
For those of you who wish you could have joined us – two options:
- We will be posting the video from the seminar on SDNCentral in the coming week.
- Or register to attend an Arista SDN event in your city. Events listed here
We also started an exclusive Arista channel on SDNCentral that will keep you up to date on Arista SDN technologies, and we will be adding other new channels in the next couple months.
Check out more about Arista Networks on SDNCentral: (our favorite is an Interview with Jayshree Ullal, Arista Networks CEO)