Subscribe / Unsubscribe Enewsletters | Login | Register

Pencil Banner

Big Switch Networks CTO Rob Sherwood on SDN in 2015: The time is now

Art Fewell | March 10, 2015
I recently had the chance to sit down with Rob Sherwood, CTO of Big Switch Networks to get his insight on whats hot with SDN for 2015.

[Rob] Probably so. So much so that we started running with the term we heard from our hyper center friends, which is a core cloud design. The idea that you have a routed core which is your ingress and egress to the data center, that lives through the lifetime of the data center. That might be there for ten years or whatever that number is. Then you have a lot of pods hanging off of it. A pod is, think of it almost like a row, but it's a collection of compute, storage and networking gear that's all certified to work together. The biggest selling point of the architecture is you've got a team whose job it is to constantly be working on the new pod design. They'll have the pod design B5 and while B5 is the latest greatest, every time you need a new project, you'll stamp out another pod of B5.

Then while the team is working on it, they'll figure out, "Okay, what is the next best version of compute, storage and networking that ties this all together? What is the next design?" Then when B6 comes in, they'll stamp that out every time they need a new project. This is kind of multi-use, multi-purpose architecture so it can have lots and lots of workloads in a pod. The real value is actually in the automation. If you were willy-nilly picking up, every time there's a new rack deployed and you've got new servers, and new storage, and new network then the automation of that is horrible. When you actually have some homogeneity, which is to say I know this is pod B4 then that's much, much less complicated to manage. That's where people are really seeing, both, the value from a purchasing cloud standpoint as well as from an automation standpoint. There's now really a huge shift in how people are building data center.

[Art] Let's say you're an enterprise customer and you're looking at deploying private cloud technology. You want to have things that are automated, and elastic, and what have you. What is your take on physical OpenFlow and physical hardware versus NVO? If I'm a customer, do I use both and they interact in some place? Do I choose one or the other? What would your advice to customers be around that?

[Rob] Well, there's two dimensions which is, one is, what is your workload? Depending on your workload, you can probably virtualize it but you may actually have some physical hardware as well, for example some database servers and things like that. Most of the customers that we talked to have kind of a 70-30 split, which is to say 70% of their workloads are virtualizable and about 30% are physical only for various reasons. Understanding your workload is the first place to start. Most people end up with some sort of mixed workload or can't know ahead of time. The answer is you would probably need both. The other side of it, it's not really an either/or decision. Which is to say, if you have and NVO solution, the packets don't match what they're getting from one vSwitch to another you need a physical network there. If you're going to need a physical network, the question is, even if the policies are maybe not that complicated, having a single point of control across the 40 switches is still easier, and cheaper, and better.

 

Previous Page  1  2  3  4  5  6  7  8  9  10  11  12  13  Next Page 

Sign up for Computerworld eNewsletters.