BY DAWN LEVY
The Internet needs a revolution -- not a mere evolution -- in its infrastructure. Scaling it up will not be enough to keep up with growing demand or to accommodate emerging technologies. Major changes are needed in the highways that convey information around the world and the backroads that carry it into homes. The National Science Foundation has given researchers at eight institutions, including Stanford University, $7.5 million over five years to investigate the technical possibilities and consequences of a once-in-a-century overhaul.
"As a group, we plan a 'clean slate' design," says Nick McKeown, an associate professor of electrical engineering and computer science at Stanford who is charged with designing the new backbone network and perhaps 100 big regional switching nodes to route communications traffic throughout the country. "This might end up as a new network. Or it might serve as a guide for how we evolve the Internet in the future."
Today, copper wire brings broadband Internet access to 10 million American homes, says research project director Hui Zhang, who leads the collaboration from Carnegie Mellon University. The researchers envision a new fiberoptic network reaching 100 million American households within the next few years. McKeown says data could enter or leave homes at 100 megabits per second -- that's about 2,000 times faster than dial-up Internet access and about 100 times faster than DSL. Over this next-generation network, file size would no longer limit what can be sent over the Internet -- removing a major roadblock to delivery of high-definition video on demand.
The new project is one of eight large endeavors funded this year by the National Science Foundation's Information Technology Research program. In addition to Stanford, partners include Carnegie Mellon University, AT&T Research Labs, Fraser Research Inc., Pittsburgh Supercomputing Center, Internet 2, the University of California-Berkeley and Rice University.
Greater capability will have a dramatic effect on daily life -- but only if the network is much more reliable, anticipates applications not yet envisioned, is economically sustainable for the long run, is easier to use and operate, and is more secure than the Internet is today.
"Although we are becoming more and more dependent on the Internet, it is scary to imagine anything really important relying on it," says McKeown, who works with Stanford collaborators Greg Watson, a senior research engineer, and Rui Zhang, a doctoral student in electrical engineering. "The Internet is not high enough quality for us to use it for most of our phone calls -- it is less dependable and the quality poorer than the existing phone network."
Today's Internet is also vulnerable to denial of service attacks on servers and viral attacks on individual machines. "This is an inherent problem that stems from the way the Internet was designed in the first place," McKeown says. "People are anonymous, and it's hard to tell where packets really came from."
Simplifying the Internet's structure would dramatically improve the quality of streaming for, say, DVD-quality television or teleconferencing. Individuals could personalize their video recordings -- "like TiVo but without home storage and with a lot more choice," McKeown says. "Imagine Netflix without the envelopes."
Now that's entertainment. But there's also economics. High-tech leaders say such broadband access could generate up to $500 billion for the economy.
But fundamental innovations in network organization and management must come first. For the researchers, the next step is designing blueprints for this near-future network by applying principles from security, economics and network research. "Part of our research project will be to design a very simple backbone network for the Internet that can provably guarantee its reliability and quality," McKeown says. What might eventually emerge from all this? "Something that we could depend on."
Stanford Report, October 1, 2003