Researchers Rebuild Their Effort to Rebuild the Internet

Article tools

Printer friendly	E-mail article	Subscribe	Order reprints
	Discuss any Chronicle article in our forums
Latest Headlines Economists Get a Lesson in Making Policy Hundreds of academic economists have tried to shape Washington's responses to the financial crisis. But many have a gnawing sense that they aren't being heard. More Coverage of the Economic Crisis Emory's President Makes a Plea for Prudent Savings Student Group Grades the Parties Vying to Lead Canada A Troubled Mexican Institution Starts to Turn a Corner A Wave of Crime Puts Universities in Mexico on Alert

By JEFFREY R. YOUNG

Reinventing the Internet isn't easy. Researchers working on a federally supported effort to design a replacement recently had to go back to the drawing board. They now think the best way to do it is to try several "best ways" instead of searching for just one — and to invite help from social scientists rather than just self-described computer geeks.

The current global computer network, born at colleges and at corporate and military research laboratories, was never intended to grow as large and last as long as it has. Some think it's already heading for collapse, threatened by the growing problems of spam and electronic attacks.

That's a problem for colleges, which these days depend on high-speed networks for just about every aspect of campus life, such as supporting course Web sites, zapping research data around the globe, and managing administrative services. What's more, many of the best new ideas in consumer computing emerge in dorm rooms, as students create programs that sometimes become groundbreaking services. (The popular social network Facebook was created by a Harvard University undergraduate, for instance.)

The hope is that the new networking project, called the Global Environment for Network Innovations, will lead to a next-generation network that can not only survive but support new tricks.

But some university researchers complained that the initial plan for GENI, unveiled last spring by the National Science Foundation and currently financed at about $5-million per year, was moving too fast. It would have quickly created an extensive nationwide test network before stepping back and deciding what kind of research questions were most important. That could have produced an out-of-date, irrelevant experiment.

"The last thing you want to do is build an enormous white elephant," said Chip Elliott, director of the GENI project office at BBN Technologies. The Boston-area firm was selected last year by the NSF to get the project off the ground. "Nobody wants to spend years of our life building something that no one wants to use."

So in March, officials rechartered (and renamed) a research advisory panel for the project and began inviting researchers from disciplines outside computer science to participate, including economists, social scientists, and theorists who study networking. What had been called the GENI Science Council became the Network Science and Engineering Council, or NetSE. The name change signals a focus on identifying the toughest questions in networking, not just building a new piece of hardware.

"We're really reaching out to the theoretical community as well as the technical community," says Jeannette M. Wing, assistant director for the National Science Foundation's Computer and Information Science and Engineering Directorate, which oversees the GENI project.

Ms. Wing and others acknowledge that the new approach means that the project will take longer to produce results. But Daniel E. Atkins, who recently finished a stint as director of NSF's Office of Cyberinfrastructure, argues that the eventual payoff could be greater.

"It may not be moving at the pace which some of the earlier proponents would like," says Mr. Atkins, who is a professor of information and computer science at the University of Michigan at Ann Arbor, "but the good news is it's being situated in a way that might have even more impact."

A Place for Radical Play

One mission of GENI is to give researchers a digital sandbox where they can try out radical new ideas in computer networking. The hope is that new approaches will improve network security, raise data speeds, and enable new features.

Network researchers often struggle to find remote digital highways to test-drive their ideas — after all, colleges don't want a researcher breaking their campus network to try out a new trick. So the GENI Project Office is working to build several virtual test tracks.

Mr. Elliott says the project hopes to build two nationwide backbones in the next six months, as well as other kinds of experimental environments, such as an urban network to test mobile devices. There's no plan for the test networks to become the next Internet themselves.

In the early days of GENI, however, some proponents hoped to quickly build a large nationwide test network. Their argument was that to truly understand whether an approach will be robust enough to replace the Internet, it would need to be tested on a large working system — one that could invite users from around the country to try it out.

Andrew M. Odlyzko, a professor of mathematics at the University of Minnesota and former director of the Minnesota Supercomputing Institute, was among the skeptics of that initial plan. "I do see the need to think about new approaches, but I don't know if actually building a real network is sensible," he says.

Researchers may still decide to build an expensive, large-scale test network, if that is the best way to tackle the questions the group identifies, according to some professors involved with the project.

Mr. Elliott points out that the new plan for the project is similar to the way the original Internet developed, in which several network operators agreed on common standards for their independent systems to talk to each other. "We know that that worked pretty well," he says.

Just what experiments will run on the test networks is still under discussion.

"There are some extremely promising ideas in the research community" for improving networking, says Ellen W. Zegura, a professor of computer science at Georgia Institute of Technology, who leads the NetSE advisory group. One experimental approach is to alter networking protocols so that networks are easier to manage. "We currently spend incredible amounts of money keeping our networks up and running," she says. More robust networks could save colleges and other organizations considerable amounts of money each year.

The Human Factor

Organizers also hope that by involving a broader group of researchers in the project, new questions will be considered. "We're the softer side of the Internet," says Donna L. Hoffman, a professor of marketing at the University of California at Riverside and co-director of the Sloan Center for Internet Retailing, who plans to go to a GENI planning meeting later this month along with several other social scientists. "I have nothing to do with the protocols or the technical details of the build-out. But we know quite a bit about online social behavior."

Taking that into account could help network engineers design systems that anticipate what users will do and adjust accordingly in advance, says Ms. Wing, of the NSF.

Those innovations are likely to be years in the making, however. Until then, "we'll have to keep patching the Internet," says Mr. Odlyzko, of the University of Minnesota. "On bad days, with chewing gum and baling wire."