Reality check: How has data center SDN evolved?

Has SDN faded, or is it being called something else?

Lee Doyle, principal analyst, Doyle Research: It's still being called SDN in the data center context. But it's bifurcated between Cisco, VMware and the white box network operating system players. Those are the three major camps you go through in the data center.

SDN is absolutely relevant, but it may be thought of in a different way and with different ingredients.

John Fruehe

John Fruehe, independent analyst: It has changed a bit. When SDN first really hit the scene, it was in the province of the big, mega cloud data centers. Companies like Google, Amazon and Facebook were using it as a way to automate a lot of their switching. Over time, we've seen changes to SDN in terms of how deeply it's getting into different product lines. We're seeing more controller-enabled configuration, as opposed to individual device configuration. But this goal of a single network driven by a single console with one software tool to manage everything is probably never going to happen -- mostly because there are too many products out in the market.

From an overall perspective, I wouldn't expect data center SDN to go away or to be a fad that dies out. It's going to continue to mature, and, down the road, this is how we'll do networking. The $64,000 question everybody grapples with is when we'll hit that tipping point where SDN is more prevalent than device-based networking. We've still got a long way, but we're continuing to make progress.

Tom Nolle, president, CIMI Corp.: Neither one of the two. The problem with SDN is sadly the problem with everything we have in tech, which is that the concept gets hyped up so far in advance of any realistic adoption. By the time something start to actually happen, it's old news.

SDN could never have happened at the data center at the pace that was expected. SDN is, in fact, progressing about the way my own surveys and model showed it would. What we're seeing today is the same realistic SDN adoption we've seen all along. This means, in the early stages, there wasn't nearly as much development as everybody who wanted to promote the idea would have liked to see.

What has most surprised you about the development of data center SDN?

Doyle: I'll answer it differently with the key things that are driving SDN. Everyone's looking at Agile, customer experience and being able to make private cloud as good as the public cloud. Those are huge drivers, as are microservices and abstracting the complexity of the network. The traditional network still requires too much manual intervention, and that's the problem we're trying to solve.

Fruehe: The most surprising thing I've started to notice is there's a clear use case in the small-business area. Companies like Ubiquiti, Aerohive and Hewlett Packard Enterprise are bringing out products targeted toward smaller business that are giving them a unified console -- that single console to manage their wired and wireless infrastructure and a lot of the other devices. It's an interesting use case, because you have a domain that's small enough where one console can effectively control everything.

Tom Nolle

Nolle: In truth, nothing has. When I did my initial surveys and projections on this, I came up with the notion that SDN was -- in the enterprise -- going to be successful largely at the pace of virtualization and hyperscale data center deployments. And that's exactly what has occurred. There's nothing new in SDN in terms of technology adoption or anything like that.

If I had to stretch, I'd say the only thing that was any surprise to me was Cisco has worked pretty hard to suppress what I'll call formal SDN, which is SDN with explicit forwarding, a la OpenFlow. I hadn't expected Cisco would be as successful in doing that as they have been. But their success in that has been offset by the fact that containers and virtualization are accelerating SDN adoption in the data center.

What is SDN lacking?

Lee Doyle

Doyle: There's not a clear architecture or a clear choice for users when it comes to the software-defined data center network. It might be clearer if you decide to go with pure Cisco, but everything else -- VMware and the white box options -- is less complete. Each vendor will give you significant benefits, but it doesn't mean a template solution. And putting the architecture together is complicated.

Fruehe: The biggest thing data center SDN lacks is that big aha moment -- the thing that makes somebody say, 'It's worthwhile for me to rip up all the stuff I have and start over with this.' It's kind of like with electric cars. Electric cars are great, and the next time you go buy a car, you may consider an electric car. But if you have a car that's running fine, and it's doing what you need it to do, you're not going to get rid of that car and instantly buy an electric because it's more efficient. It's just one of those things that's off in the distance for you to think about when the time is right. That's the biggest challenge SDN has. It has an install base that's hard to battle against, and it's hard to overcome that.

Nolle: Truthfully, it's lacking a realistic understanding and mission. Right from the start, it was obvious what SDN was good for and where it would be useful -- the data center and data center interconnect in one thrust and optical grooming in the other. But we've had SDN stories about almost everything you could do with SDN, except the stuff that was practical.

That's our real challenge right now. We need to make people understand SDN is doing exactly what SDN was intended to do. It wasn't nearly as exciting and transformational as everybody hoped it would be. However, it is incredibly useful, and it will become more useful over time. Realism is what we need here.