Bell Labs' Weldon on Infinite Bandwidth

By framing the problem correctly, research lab is focused on making bandwidth seem infinite even it if can't be.

April 9, 2015

7 Min Read
Bell Labs' Weldon on Infinite Bandwidth

At a time when Bell Labs celebrates its past Nobel Laureates with a new garden at its Murray Hill, N.J., facility, the organization's head is looking to the future.

Specifically, Bell Labs president Marcus Weldon is looking forward to applying Bell Labs' innovation methods to industry challenges -- specifically, massive bandwidth needed to support mobile apps and devices, including the Internet of Things.

In a rapid-fire interview with Light Reading, Weldon, who is also Alcatel-Lucent (NYSE: ALU)'s CTO, also explains why having 700 people is enough to get the innovation the industry needs, and how those people need to work together. He also details how Bell Labs has been reorganized so that each researcher chooses to work on one of 13 "Future X" projects, of which mobile bandwidth is just one.

Figure 1: Bell Labs President Marcus Weldon Weldon spoke earlier via video with Light Reading Editor-in-Chief Ray Le Maistre and you can see that here. Weldon spoke earlier via video with Light Reading Editor-in-Chief Ray Le Maistre and you can see that here.

On framing the problem of getting enough bandwidth: "The seemingly infinite capacity is our goal, it's our charter at Bell Labs. Stated that way, it is a typical Bell Labs problem. If you just said, 'I want 10 Gigabits per second' or one gigabit per second, that's a problem anyone could state and everyone could attempt to solve. The right framing of the problem is often key -- when you frame the problem correctly, you build the right solution."

On solving the problem of delivering seemingly infinite capacity: "To get to seemingly infinite bandwidth capacity, yes, you'd use more radio capacity, more spectrum, but you'd use all radios at once, and you would optimize how flows flow to each radio depending on the needs of the flow. If it's a very high capacity flow, you'd use a high capacity interface, if it was a highly mobile flow, you'd use an air interface that had very good coverage as well as capacity, if it was a local flow you'd use a local air interface etc. It would have the idea of guiding the flows depending on their characteristics, which isn't just bandwidth and latency.

"And of you course you'd try to do that in the most economical way and do as much as possible in the cloud and leave as little processing as possible in the outside network as the outside network is going to get closer and closer to the end user with small nodes and small cells."

For more on the technology shaping the wireless future, visit the dedicated 5G section here on Light Reading, and register to attend the upcoming "Building America's 5G Ecosystem" event in NYC.

On what could be involved in solving that problem: "I think there are three or four dimensions: One is I need to go to smaller and smaller nodes and those nodes will be disconnected from each other, because they'll be everywhere. I need to connect those nodes together so I need a new backhaul architecture that doesn't require me to run fiber everywhere and can be adapted. So we are looking at massive MIMO or beam-forming for backhaul, because I can really put my node down wherever it is, my small cell, and it will self-align to the backhaul. So the backhaul is through a massive antenna array that is forming beams to wherever these small cells get deployed. They don't have to be situated at specific points, essentially the network auto configures the backhaul.

"You don't want to it to be micro-antenna links to each one because those are expensive and they have to be 80 feet off the ground. You'd rather have it be using low frequencies and doing backhaul over low frequencies and then doing beam-forming to direct the beams just to the small cells and that's the most efficient way to use the spectrum.

"And then I want to do millimeter waves on access layers so I get access to multiple gigabits of spectrum up in the millimeter wave frequencies, such as 60GHz or 30GHz, around there because there is lots of spectrum up there. So I want to find ways to use that spectrum that has very distinct properties -- it doesn't go through walls, gets absorbed by air. That's why it doesn't get used very often. So I have to build an architecture that works with that and my low-frequency spectrum.

"And then we need a new air interface. If an app is just sending a small amount of data every hour, then that looks like a machine in terms of traffic properties and we need to build an air interface for that. And then we need a control plane that makes all of that work together -- that is sort of the big problem.

"So: beam forming, millimeter wave, a new air interface for short burst communications and a control plane that guides the traffic over all of those resources optimally."

Next page: Prioritizing the challenges to infinite bandwidth

On prioritizing work on these challenges: "Our Future Wireless project has 11 subtopics. We have to innovate on each of them and together. You can't solve one part of that problem without knowing you can solve the others, otherwise you wind up with an incoherent solution. The 11 subtopics are actually just in the Future Wireless project. We have 13 Future X projects, of which this is one that has subtopics around beam forming, millimeter waves, a new control plane and a new air interface. You can see the breadth of our scope. We have these 13 big topics and within each are subtopics that are the ingredients that come together to create the solution."

On prioritizing research efforts: "People are our main expense. What is the minimum number of people you need to innovate in all the dimensions you'd like to innovate?

"There are 700 people in Bell Labs -- is 700 the wrong number for that? I don't think so, I would argue that for each big innovation area, you need only one or two original inventors. If I want to innovate in 20 areas, I might need 40 people. I need tens of people to then implement that and challenge and innovate to create a solution, but it's not clear to me that 700 is the wrong number to innovate in all the dimensions of the network, because there is no project I'm not doing that I would like to do.

"So yes I am not working on materials chemistry of new biological materials but we are working on neural network architectures and modeling network control planes after some aspect of the brain. So I would say I can do all the things I want to do if communications, collaboration and connectivity are the things I want to work on. Then I have all the resources I need. Of course we need to work harder."

On bringing focus to Bell Labs' research efforts: "When I took over Bell Labs, I found that there was roughly one research project per researcher and since I have 700 researchers, that would be 700 projects. And they weren't really coordinated particularly closely. So the Future X projects, of which there are 13, are ways to couple those together and encourage researchers to work together and create something bigger than their own.

"We've done it organically by just encouraging them with a compelling vision about, look, we need to go to this much in optics, we need to reinvent data center networking, we need to reinvent wireless. By making the statement compelling enough and ask researchers to get involved, we've found that organically people have flocked to the projects. Each project has 10 to 20 to 30 people and we've put two leaders in charge. Researchers are actually naturally collaborative with a bit of encouragement."

— Carol Wilson, Editor-at-Large, Light Reading

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like