AT&T Labs preps for 10x increase in national core throughput
with Krish Prabhu, president and CEO of AT&T Labs
In late June, AT&T (NYSE:T) named Krish Prabhu as president and CEO of AT&T Labs, the company's research and innovation hub. Prabhu, an engineer by training, is an industry veteran with 30 years of experience in both the technology and business sides. FierceWireless Editor Phil Goldstein recently spoke with Prabhu about what AT&T Labs is doing to help AT&T Mobility innovate in its network design, improve wireless signaling for smartphones and launch LTE.
FierceBroadbandWireless: AT&T Labs has several key research areas. How do you think research into networking and software systems is going to evolve, particularly as more strains are put on wireless networks from mobile data?
Prabhu: There are a lot of different things that are happening in network design. Mostly with the networking we're talking about is large networks, and with software systems, it's managing large software databases, literally trying to find a needle in a haystack. When we look at things today that use the power of search engines and looking at the power of clustering of data, [and] a lot of the advances in software that have happened over the last decade or so because of the Web players that really pushed the frontier, a lot of that sort of technology can be mapped and adapted to running the network more efficiently.
Let me give you a couple examples that highlight this. If you take a look at our wireless network, wireless data traffic has grown by 8,000 percent over the last few years. As we look at these applications behind platforms that are impacted by how the network works, in many cases they take what happens in a wired network and just adapt it on a mobile platform, which kind of captures to some extent some of the limitations and the capabilities of the handset platforms you are designing the app on, and certain assumptions, primarily on the basis of what the standards allow for the handset to talk to the network.
So in that context, when we look at flow control, what state the handset supports, what happens when a handset gets flooded with return data coming back from the network, there is a lot that can be done in making sure that the device is optimized to work with the network well within what the standards specify. In the case of voice, it didn't really stress the network. You just followed the protocol and the call went through, though there were slightly different technologies. But in the case of data, it's very, very different, especially when you look at video. There is a just a whole lot there to [manage] between the web server at one end and the handset at the other end, with the network being in the middle. So there is a lot of research into algorithmic flow control that we are working on right now, experimenting, interacting with app developers as well as the website guys. If you just look at mobile, 50 percent of the traffic is video today, and it's only going to go up. And it's very much in our interest, both the network guys as well as the website and the handset guys, to make sure we get the most out of the network.
FierceBroadbandWireless: How involved has AT&T Labs been in working with the larger company's network architecture team on innovation, and how do you expect that to change?
Prabhu: The network architecture is designed not to be too innovative, as you can understand because, a) it needs to conform to standards, and, b) when you do architecture you are really planning for long-term growth and making sure whatever you put in the ground today doesn't have to be ripped apart, because you don't have that option. So from an architectural standpoint, we have certain guidelines that we follow to make sure that our network design is consistent with the architecture.
Now, within that architecture we have a lot of room for innovation, and a lot of room for making sure we do the right things. As a case in point, when you look at our evolved packet core, for example, and you look at the throughputs through the gateways--not just for today's traffic but traffic projections looking forward four or five years from now--video is going to be the predominant component of that traffic as we go through to 4G. We work closely with our vendors, we look at elements of policy control, policy enforcement, end-to-end video streaming rates and throughputs. There are lots of different parameters that we really have a lot of latitude on and can specify. We are really involved and do it in a collaborative way with the vendor community.
FierceBroadbandWireless: Will AT&T Labs be focused on innovations in things like signaling and antenna design, which aren't very sexy but are actually pretty crucial?
Prabhu: Oh yeah. That's a great question. What I have talked about up to this point is really layers one through three. It's the lower layers, and it's all of our traffic and it's all about bits moving back and forth. But when you look at layers four and above in the architecture, it's all about using an IP backbone and how do you implement the intelligence you need to set up sessions, how do you handle wholesale HTML traffic when it hits the Web, how would you set up a services framework that makes the network application-centric, how would you open up those applications to well-defined APIs for third-party developers.
The bulk of our Labs activity is focused on that, because the essence of this network, especially in the 4G network, is really all about a wholly different ecosystem and architecture. It's taking elements of what has served the non-mobile Web in the last decade and transitioning that not the mobile Web. And when you look at the problem in that context, there is a lot that has happened in the higher layers, and most of our research is focused on that.
FierceBroadbandWireless: And I know signaling, in particular, has been a cause of concern among vendors in terms of determining how often and how efficiently smartphones and other mobile devices interact with the network. What do you think can be done to improve signaling so that it's not putting as much strain on the networks?
Prabhu: When you look at signaling on the phone, the phone becomes aligned with the IMS [IP Multimedia Subsystem] architecture. When you look at LTE, for example, voice over LTE and video over LTE will be carried in an IMS framework, and that will be, in many ways, the first time we will see an IMS implementation. IMS actually does specify not just the signaling sequence, but also the extent of signaling you will build. What we can do within this framework is to make signaling very efficient--we could do a lot in the implementation aspects of it.
So, for example, in IMS, we have three main elements. There is a session control element, there's a border gateway element, the so-called border controllers, and then there's the driver databases that have all of the preferences. So fundamentally, as you look at session management related to these three elements, by using really state-of-the art protocols like diameter signaling between these elements, having intelligent routers specialized to do diameter signaling, looking at the design of the network to optimize it for higher throughputs, there's lot of details when you look at it at that level, that will make the signaling very efficient.
What's happened today, to some extent, is that in a 3G world in which signaling was largely done to facilitate voice, and when it ended up carrying a lot of data when the smartphones came, this really stressed that to the limit. In the LTE world, within an IMS-compliant signaling platform, we will be able to correct that.
FierceBroadbandWireless: What role do you think AT&T Labs will play in AT&T's deployment of LTE and then especially later when we move to LTE-Advanced?
Krish Prabhu: AT&T Labs is very much in the thick of things as we move toward our LTE launch. The Labs is very much involved in not just working with our vendors, but also it's involved in the design aspects of the network. It's involved in understanding how you get maximum coverage between 3G and 4G networks. Part of the Labs works closely with the handset and device manufacturers to certify them for use. So there is a ton of activity in the Labs pertaining to LTE. A big part of our LTE rollout, especially as you get to voice and video, is indeed the IMS compliance platform. We are involved in virtually every aspect of the LTE rollout.
FierceBroadbandWireless: What do you think is going to be the biggest challenge for AT&T Labs in the years ahead and how do you plan on tackling it?
Krish Prabhu: There are multiple fronts. Maybe time will tell, and when we come out on the other side, we will be able say definitively what is the biggest challenge. But right now, it's handling the capacity issues, and making sure the customer experience is the best we can offer for a given capacity. I talked about flow control and looking at ways to minimize stress on the network, which ultimately affects the customer experience. Just to give you a sense how difficult the capacity issue is, we have a national core and we have 11 or 12 large metro regions across the U.S. And today the national core has a throughput of two to three terabits per second. Over the next five years, our traffic volumes tell us that when we launch LTE each one of our 11 regional cores will have a throughput of two to three terabits, and the national core will have a throughput at least 10 times that. ... We are very involved in the solution to that problem. We've identified a layered approach to get us there even as we support the launch of our LTE network and get LTE to 90 to 95 percent of our end-users. That to me is the biggest challenge.
The second biggest challenge is making our network open to an application-centric architecture, and third is developing an ecosystem of application developers who will leverage the true power of the network. If you look at the application development environment in the fixed Web, you'll see what a rich ecosystem it is. That whole thing is going to be transitioning to the mobile Web, which in a way is a challenge for all of us.