Video Meet OpenDaylight Project Executive Director Neela Jacques (Video) 14
Today's interview is with OpenDaylight Project Executive Director Nicolas "Neela" Jacques, who has held this position since the project was not much more than a gleam in (parent) Linux Foundation's eye. This is one of the more important Linux Foundation collaborative software projects, even if it's not as well known to the public as some of the foundation's other efforts, including -- of course -- GNU/Linux itself.
Neela Jacques: The OpenDaylight project is at this point the largest and most successful OpenSource project in the networking industry. We were created just over two years ago because there was a problem in the software-defined networking industry where there are about 30 different controllers out there more being created every day. They're all proprietary. And the industry just didn't have a single platform to write their apps to. What we found was that end users weren't deploying software-defined networking and really there was just this large need for a platform. And so the IBM brought the entire industry together to build this platform.
Robin Miller: IBM is definitely a good daddy to have because if they use it, people use it. So how industry after two years only how many people or companies are moving to software-defined networking?
Neela Jacques: Right. Software-defined networking as a trend has really been gaining tremendous momentum. In fact, AT&T just shared last week that they put an SDN based solution on the market called Network on Demand, which they've built on top of the OpenDaylight project. We've got about 460 developers who contributed to the project, about 20 different products out on the market that are based on OpenDaylight. We've got end users like Orange, Telco in Europe, like Comcast here in the United States. The Large Hadron Collider, the particle physics folks, as well as just a wide range, things from connected cities in Bristol to some enterprises using us for OpenFlow. So a wide range, I don't have quite a number for you.
Robin Miller: It’s okay.
Neela Jacques: But it's been pretty impressive.
Robin Miller: I think we can quantify it as a whole bunch, isn't that the scientific term?
Neela Jacques: Yes, I think that's true. What's interesting about SDN is there's really three emerging used cases that we see. The first one is actually the least sexy of them all. People's network the way that they are today is just too manual, you have to make manual changes everywhere, it's not programmable. And so there's a basic level of being able to orchestrate, to automate the networks that you have, and the truth is you can't throw out all the gear that you have and buy a new gear to be able to do it. I mean there are all these people who got this, oh, we've got a perfect solution for greenfield.
Robin Miller: Yes.
Neela Jacques: Especially on the telco side, they've got a year old equipment that's still working, they don't want to throw that out, and so I think that that's the first piece and understanding as a result, we need to be able to work with a wide range of hardware. It's not just going to be OpenFlow, you're going to need to work with BGP PSAP, you're going to need to be able to use NETCONF. And so I think as end users begin to deploy SDN on to their existing infrastructure, some of what we need is the hardware manufacturers who’ve been playing it a little bit fast and loose with some of the standards out there.
Now we're going to have a de facto standard in code to make sure that they work well with, and I think that's one of the first things that we need to do to really help nail this orchestrated programmable management of the networks that we have today, so that's use case one.
Robin Miller: So what you're really saying there is as much as anything else, you are now acting as a standard spotty which is necessary and unsexy.
Neela Jacques: Yes, well, I would say a de facto standard, so in the past
Robin Miller: Well, standard de facto, de-standard, however doesn't matter.
Neela Jacques: Exactly, in the past what would happen is you'd get an IETF draft and every hardware manufacturer would say, do I follow the letter of this law. And so in a sense we're going from a written law to now a very rich case law where we have software, you can test your hardware against that software.
Robin Miller: Oh, yeah. Again, it's a de facto or not, it's a standard, we know it's going to work, we know you can plug it in, and therefore it's highly salable to the vendors, and highly usable for the users, okay. And that’s number one, the unsexy, the dotting the Ts crossing the Os or the zeros for Europeans and all that.
Neela Jacques: Yes.
Robin Miller: And what's less sexy or more sexy, number two.
Neela Jacques: So the number two and we're seeing tremendous excitement around this is NFV, Network Functions Virtualization.
Robin Miller: Okay.
Neela Jacques: So this came out of the Telcos, about two, three years ago I think it's probably closer to three. They realize that one of the key problems with their network was the fact that they were tied deeply to the physical, and so they said what we really need to do is make sure that when we go ahead and spin up a load balancer when we need to add more users on the system, whatever we do we want to add VMs rather than add physical hosts.
Robin Miller: Yes.
Neela Jacques: And specifically, it's not just that the VMs are better and cheaper and easier to run. It's actually that once there are VMs, I can do a whole set of things, I can think about what order I put those VMs in; that’s something called Service Function Chaining. I can make it that I can use the bandwidth on my network better. So if I know that it's not – if it's New Year's Day and everybody's call on their mom to wish them Happy New Year, well I don't want to run anything else on that network.
But if on the other hand it’s lower down, I can use that for things like backup, I can go in and do syncs between parts of my network, so I can use the network in some more intelligent ways.
Robin Miller: Very good. Also, even just on a more basic thing of just idiots running websites, I remember when some years ago Slashdot was totally dependent on one load balancing switch, lots of servers. But we had one big load balancing piece of hardware and one day it stopped working. It was a single point of failure and we all knew it. And we had some very smart network guys and they were saying that sooner or later it would fail, by God, and it did. I remember standing around with some of the guys, and we were thinking, gee, wouldn't this be better if it was software controlled? Instead of hardware?
Neela Jacques: Right.
Robin Miller: So we were wishing for SDN 10 years ago.
Neela Jacques: Right. And there's so much. What's interesting is people talk about telcos as dumb pipes. And the truth is, part of why they've been dump pipes, is it's been so hard for them to innovate because their network has been built over 20 years and it's got to keep running, we all depend on it. They don't have the ability to shut down for three months and rebuild a network from scratch. And so, you know, if we move to a model where it's defined by software rather than hardware, where they can make changes, there's actually a tremendous amount of opportunity in smarts, right. Right now the sound quality or the video quality wasn't good, I would love to be able to get a message and pay $5 and suddenly get exactly what I need. If I think about walking around the streets, wanting to get the ability to get information that appears on my phone based on my location or based on other media, actually the amount of things that they can do could be huge if their underlying infrastructure wasn't restricting them from doing it. And right now it takes them 6 to 12 months to even trial out some new crazy idea; even if only 5% of ideas succeed, you now have a problem which is it takes 15 years and all these over the top guys are coming in and they're able to drop in innovations like this.
Robin Miller: Sure.
Neela Jacques: Anyway, that was number two. Use case number two was network functions virtualization.
Robin Miller: Yes.
Neela Jacques: It is very important; we're seeing every major telco go there. There are many parts of it, it's the mobile part of the network, as well as the fixed part of the network. This is a trend. It has taken off and people like AT&T, I think are ahead of others. But we're going to see over the next 10 years massive amounts of change in opportunity because of this underlying shift in the infrastructure.
Robin Miller: Absolutely.
Neela Jacques: So the third one is cloud based infrastructure, as you obviously know that more and more these days, if you're creating something, you're going to house it in a central location. We've moved from on-premises software to hosted software. The problem that we have is how do we provide multi-tenancy? And networks were inherently not built for multi-tenancy and so what we've seen people do in a virtualized world is use something called a VLAN, you're probably quite familiar with a VLAN. It sounds like it’s a virtual local area network, but it isn't that at all. In fact it is the instantiation of a physical network, just in the virtual world. But it does not do a very good job of actually keeping the traffic separated where it needs to be separate. It is time consuming to set up and test and in general is far too manual, so we know that we need to move to a more virtualized version of the network. OpenStack is becoming the de facto standard architecture for the cloud.
Robin Miller: Yes it is.
Neela Jacques: Yes, OpenStack doesn't have and doesn't come with an STN controller. And so the initial model of OpenStack is, hey we'll talk the ML2 directly down to layer 2, which is really anathema to what the cloud is all about. And so connecting on one hand, a rich open SDN control like OpenDaylight with a rich cloud orchestration platform like OpenStack just makes all the sense in the world. And so this is certainly a huge next step that we're moving as part of that pieces of neutron which is their networking, the networking stack they have has to improve and we're helping them do that. And then of course we have to be able to take information from their API, the neutron API I just talked about and being able to instantiate networks in a multi-tenant way for the users.
Software defined networks .. (Score:2)
Secure? (Score:2)
OpenDaylight written in Java .. (Score:2)
Googling this guy (Score:2)