(Photo: Neela Jacques, right, and Jim Zemlin.)
The OpenDaylight Project has named Nicolas “Neela” Jacques as its executive director. He’s the first person permanently filling the spot, which has been temporarily held by Jim Zemlin, executive director of the Linux Foundation, since OpenDaylight was created in April.
Jacques was selected not for his open-source expertise (of which he has pretty much none) but as someone who knows the networking industry and wants to rally people around the cause. He’s joining from VMware, where he was one of the first employees to push the importance of cloud computing, back in 2007.
Jacques and Zemlin stopped by the SDNCentral office last week to talk about their jobs and the future of OpenDaylight, which, is getting ready for Hydrogen, its first code release, on Dec. 9.
Are you still working at VMware?
Jacques: No, no, I’m giving up the comfortable job.
Wow. So, OpenDaylight is your full-time gig. You’ll work at a dot-org and wear sandals…
Zemlin: We dressed up today. I’d be in sandals.
Jacques: This is a career shift for me. I’m very excited. I didn’t wake up one morning and say I’ll be an open-source guy. It sort of came from the other direction.
Seven-and-a-half years ago, I joined VMware because I thought that virtualization could, in some sense, change the world. The further I got, the further I realized that in order to have an impact on IT and the data center, it wasn’t enough just to virtualize compute. And so, I was one of three or four people who came up with the idea of the software-defined data center. It was my job, internally and externally, to go around and lead the education about that fully automated data center and the virtualization of not only compute, but networking and storage. Going out to customers, what I heard over and over again was that if you just do the other two, and you don’t virtualize networking, then you don’t get the benefits of a fully automated data center.
Here’s my biggest struggle. VMware buys Nicira, and I see this whole group coming up within VMware looking to deliver a proprietary network solution.
The challenge is that unlike compute virtualization, networks are all connected. With compute, you could pick off a small piece and virtualize that and see if it works. With networking, it’s very hard to go to a large corporation and say, “Let me network just these two computers, and everything else I’m going to leave alone.”
Not to say anything negative about VMware, because there’s nothing negative to say, but: The entire industry has agreed that SDN is necessary and is going to happen. The big question is how and how quickly. What I was thinking of already, before I met Jim, was that what people really want is an open platform, one common platform that they can plug everything into. And what you get in the industry is that everybody’s got their own solution, and it’s a wonderful solution as long as you buy every component from them. But no one actually makes every component, so, from a customer standpoint, where they’re struggling, is that they’re having to buy two or three SDN solutions, minimum, and they’re being pitched 25 or 26.
OpenDaylight has had its share of controversy. What did people think when you told them about the job?
Jacques: As I went through and did my due diligence, it was interesting [to see] the amount of passion people brought into it, positive and negative. I met people who were sure it was going to fail, and they absolutely wanted me to hear all of their reasons. You had people who were pounding the table: “It is gonna fail! It is!” But you had other people saying, “This has to succeed. I don’t want to live through another 20 years of the same darned networking model.”
What exactly is your job?
Jacques: My job is to facilitate the success of the project. There’s a big difference between collaborative development and an enterprise software company. The enterprise software company is command-and-control. The open-source model is fundamentally different. I work for the developers more than they work for me.
In fact, they don’t work for you, right? They can flip you off and walk away at any time.
Jacques: (Laughs.) Absolutely. Once you realize that, you start to ask yourself: How can I bring in my experience —and some degree of leadership, obviously — to help support this community of people? The first big thing is helping them prioritize. Bringing end users to the table.
The first seven months [since OpenDaylight began] have been about code. Now that we’ve got a release coming out, in December — now’s the time when we can go out and get the guys who spend their life running big data centers and say, “Here’s what we’ve got. We’ve got a set of code, and you’ve got problems. How does it map?” Some of them will say, “Awesome. I’ll take it.” I think a lot will say “This solves this problem relatively well, but I’ve got this other issue.”
For example, the carriers right now have been struggling with this issue where it’s incredibly expensive and complex to keep delivering all these pizza boxes for all these different services. What they really want is to have all the network functions virtualized. They’ve got a clear need, and we, as an organization — NFV [network functions virtualization] is part of our charter. NFV and SDN [software-defined networking] are the two major things. Now is the perfect time to get those carriers involved with us. So, we’ve started an end-user group. My job is to make that successful. Help create the structure and support. So, that’s an example of an immediate task.
Zemlin: It’s really analogous to the role I play in the Linux Foundation. For example, where there are resources needed to develop a test framework for some component of the project, I’ll go out and get the resources and funding, and the people to get that done if that’s not being done through volunteer work. The developers make the decisions on the open-source project. That’s the easiest part of the job.
Jacques: The challenge of this model is that you actually put the developers in front of the end users. There’s an art to being able to get that to be constructive, because it’s easy to have some end-user rattle for 20 minutes about some little feature that they like.
Zemlin: The roadmap is the same as in roadmap in Linux: There is no roadmap.
We use the term “weather forecast” for “roadmap,” because we generally know where things are going, but you don’t want to stifle unexpected innovation. For example, who knew somebody would submit a DDoS application to OpenDaylight? I certainly did not.
Jacques: We started with centralized control. It’s a large problem, it’s immediate, and it’s tractable. That controller doesn’t try to do everything, but it does have to enable it. So we have the northbound interface. There also has to be a management-stack element. The controller is getting commands from something; this provides that output.
Are you sure it’s all going to work together?
Zemlin: When you look at some of these components in a typical product development cycle, you can take a lot of shortcuts architecturally to get that thing out the door that you might regret later. In an open-source project, that’s almost impossible to do, because the nature of the collaborative development is one that promotes pluggability and stability. The phone calls I always get about Linus Torvalds ranting tend to be about breaking the APIs in Linux.
How do OpenFlow and OpenStack fit into all this?
Jacques: The ONF [Open Networking Foundation] is seeing there’s this battle going on out there, and there’s really some value to standardization, and you can see the value of OpenFlow 1.0 and 1.3. They’re about delivering the standard, and we’re about delivering code.
If you look at OpenStack — if you’re building a data center, or if you’re building a cloud stack, it’s not just about networking stack. It’s about compute. It’s about storage. So, this plugs in very well with OpenStack. In fact, we have some developers who contribute both to the OpenStack networking stack as well as to the OpenDaylight project.
Now, suppose — and I’ll pick on OpenFlow, only because it has a name — suppose some alternative to OpenFlow shows up, and some faction wants it to be in OpenDaylight? What then?
Zemlin: You’d have to ask the developers, I suspect.
Jacques: It depends where in the life cycle a technology is. Early, it’s useful to have multiple things so that the market can choose what the winner is. Longer-term, it’s useful to have a standard, because the more different options you have, the more complex it is. I expect you’re going to see both models. You may see times when the developer and user community say, “Give us choice.” Other times, they’re going to say, “I don’t want to choose among five things. They all basically do the same thing. Let’s pick the dominant one and make sure everybody’s using it? ”
I’m still wondering how you’ll keep OpenDaylight from just expanding forever, spreading like a liquid?
Zemlin: I’ll use an example of open-source that has been criticized in the past: Open Office, where you have this big, monolithic code base. A few years ago, it would take hours, maybe half a day, to compile, right? You don’t see that here. My sense is that the architecture will indeed come together in a meaningful and clean way, just as other similar projects have.
Jacques: And wouldn’t you say, Jim, that in the open-source model, there’s a balancing mechanism for that? Which is the ability to fork the code. If I look at Linux, is Android hurting Linux by taking it out of scope? We all know what Linux is for, right? “Linux is for workstations — I mean, servers — I mean—” You know, if someone wants to work on it for cars or microwaves, if there’s a defined use case and people who find code that they can leverage, those things can come together. If that’s a separate thing, it can become completely separate.