After tracking him from VMware to Big Switch Networks and then to Cisco, we thought it might be nice to talk to Howie Xu — now a senior director in Cisco's Cloud Networking Service Group — about something besides changing jobs. Turns out he'd recently started looking into OpenStack, approaching it with a point of view steeped in networking and virtual switching. Here's a bit of our conversation about OpenStack's maturity (or lack thereof) and what role the platform plays.

You mentioned that going to an OpenStack Summit changed your perception of OpenStack. How so?

Xu: In the past I was thinking: OpenStack, maybe that's an open-source solution comparable to VMware's vCloud. But I changed my mind after talking to half a dozen customers closely on why they are using it.

If we compare OpenStack to VMware's vCloud suite, it's very much like comparing NoSQL to an Oracle database. It's not like Oracle was designed from the beginning to deal with this super, scaled-out, database. If I need a super scaled-out database, I don't really have a commercial choice at all.

Same thing with those customers I talked to who started using OpenStack. It's not like vSphere or vCloud was even a choice. For the scale, the super-scale deal we're talking about? Their alternative is to write everything from scratch. And with OpenStack, you'd also have some people to support you in the community. There is a roadmap, and there are lots of passionate developers talking about it.

So that's how I sort of look at OpenStack these days. I feel like yes, it's nowhere close to being mature, but it's solving a fairly unique use case that no commercial product is able to do.

It's also free, right? Is that what first gets people attracted to it?

Xu: A lot of that is not about necessarily saving money. I don't really have another alternative. If you look at their cost, if you add a dev-cycle cost, it's actually still a lot.

What goes into that cost?

Xu: Let's say you pay a VMware license. Then you don't have to pay humans to develop code for you. And you have peace of mind that it's going to work, rock-solid. So why do you even have to develop your own code? That's crazy, right?

Here, [with OpenStack], you don't really have a commercial solution. You have to develop all of that by yourself. So even though in OpenStack, the stack is free, you still have to hire good engineers, pay them well, and retain them well. Hiring a good OpenStack genius is not easy.

How hard or easy it for an engineer to get into OpenStack?

Xu: I think there are two different pieces here. One piece is just the coding experience — Python. I wouldn't say that's easy, but there are tons of people who sort of have that skill set. The other aspect is actually the workflow, the customer experience, the understanding of the use cases. Understanding the architectural trade-off, you often have to make for a scale-out solution. So, for that, it's more difficult.

It sounds like an experiential kind of thing, almost. You just kind of have to start doing it and build up knowledge that way.

Xu: Yes, and if you want to use OpenStack, it's easier than adding additional feature stacks, fixing bugs. But because of the current state of OpenStack, you don't have a turnkey solution. It's not as if someone can just give you a piece of stack and then you can be worry-free. You also have to hire a developer.

I remember a few months ago when I started using OpenStack myself, I literally had to Google a kernel patch in order to use OpenStack. That's sort of where we are. Even if the code were to get more stable, get more feature sets — today the reality is that there is no installer, even. There's no way to easily monitor what's going on. So it's not even comparable to VMware's solution, the hypervisor solution, that's just a few clicks to get started.

But a lot of new things are happening. One good thing I've seen at OpenStack conferences is that there are so many passionate engineers trying to improve it and who are working hard on it. I think that this is one thing that gives me some hope. Havana is better than the previous release, and IceHouse [scheduled for April 17 release] is going to be much better than Havana. There are a lot of new things happening in this space. It's a pretty exciting space. Just be careful in what you're getting into.