The future of networking

We recently visited BBN Technologies, the storied Cambridge, Mass. R&D outfit, where a lot of smart people get to work on some really cool stuff, a lot of it generated by U.S. Department of Defense interests, priorities, and of course money. One emerging area in wireless networking is the concept of “content-based networking,” says Jason Redi, principal scientist with BBN’s Network Technologies group. “You might ask ‘show me the maps near where I am’ without having to know where those specific files actually are,” he says. It’s related to the idea of intelligent caching, in effect offloading some aspects of communications to an intelligent storage infrastructure. Akamai does this as a specialized overlay to the network. True content-based networking will be an integrated part of the network’s behavior and intelligence, Redi predicts. Some of these concepts are emerging for the DoD’s Wireless Network After Next (WNAN) project, part of BBN’s research portfolio. The goal is a low-cost, highly-capable tactical radio to enable reliable communications with every soldier and every device at all operational levels. But that entails advances in ad hoc wireless networking, especially in opportunistically evaluating the RF environment and adapting to it, and in relating information requirements to location and other tactical variables. Another area of Redi’s study is ultra-low power ad hoc wireless sensor networks, a project called JAVeLEN. The project has created a novel design that can use 99.7% less power than industry standard protocols. At the same time, the JAVeLEN network is more reliable, handles much greater node densities, supports mobility, and performs well under varied loads. The network’s intelligence keeps RF receivers off when not in use and coordinates how and when the nodes listen to each other. The nodes can find each other automatically, and automatically build their own routes through the mesh. One key area of development is optical networking, where the goal is to move beyond the physical layer to add much more intelligence to the upper layers of the network stack with new protocols. BBN is working in this area with UCal Davis, UCal Santa Barbara, and Verizon among others, according to Kristin Rauschenbach, BBN’s director of optical networking. To dramatically boost throughput, beyond simply playing with the lasers and fiber, requires “lots of internal cooperation” among the various protocol layers, she says. If that sounds like optical switching, it is: being able to change the wavelength and switch from fiber to fiber, assigning traffic to the different wavelengths. “Today, optical switching has very limited capability: it moves light beams around,” Rauschenbach says. “It can’t do anything within the ‘pipe.'” The future is about “making intelligent decisions about moving traffic,” Rauschenbach says. “This intelligence is what lets you improve efficiency.” The result? Bandwidth on demand, as part of a dynamic, constantly changing optical network fabric. Verizon has just announced a bandwidth-on-demand service for large commercial customers; the demand triggers the activation of the service. In future, based on the kind of research BBN is doing, the optical network will be able to increase and decrease bandwidth as needed. One example of where these concepts are being applied is the CORONET program, funded by DARPA. The focus is on creating ultra-fast service set-up and tear-down, with fast and efficient recovery from multiple network failures, and rapid network reconfigurations in response to changing bandwidth demands. Those demand changes can be in response to what Rauschenbach calls “flash crowds” — a sudden, unpredictable, and big demand for network bandwidth from a given region or even a locality. Demand could be generated by a disaster like the September 11 attacks or by a developing hurricane threatening the Gulf coast or more locally by something like Boston parade celebrating a Red Sox World Series victory (whenever). Today, such sudden demands either can’t be met, or can only be met by creating oversubscribed networks. By contrast, an intelligent optical network would be able to throw more bandwidth automatically and dynamically to a region with a critical need, Rauschenbach says. Another area of research is Project GENI, directed by Chip Elliott, chief engineer at BBN. Funded by the National Science Foundation, GENI is sort of like a massive digital electronic Erector Set for playing with the next iteration of the Internet. The first stage is under way: paying out $12 million in grants to bring together existing research facilities into a coordinated, interoperable whole. Elliott says the initial GENI infrastructure will be “up and limping” in the next few months “so people can [then] start arguing constructively” over the future of networking. Elliott says early prototyping of some projects will start to appear in early 2009. GENI is an “experimental infrastructure” that will bring together computation, storage and communications. A main research focus is how virtualization can be exploited in all three areas. Elliott foresees “end to end virtualization” of all these elements — computation, storage and communications, coupled with “process migration” — being able to create, move, expand or contract, and close computational tasks seamlessly. “Today, networks change slowly,” Elliott says. “Virtualization lets you change this stuff much faster.” Such dynamic networks will be much harder to penetrate by attackers. And network functions can shift to servers and routers as loads diminish or rise. One variant of this kind of fluid networking is reducing electrical use during off-peak periods, by powering down gear that’s not needed to meet traffic demand. One corollary is integrating into the virtualized core vast new sources of real-time data drawn from wireless sensor networks.