Carnegie Mellon has just taken $7.5 million NSF money to look at ways to build the next generation of the internet. The article here includes some interesting quotes from Carnegie Mellon professor Hui Zhang, going against the popular theory that the internet can continue to scale the way it’s been built – saying that the complexity of growing this system and keeping it running are becoming overwhelming.

A new and crucial chapter in the history of the Internet began last week. Expect all sorts of evolution vs. revolution battles before the chapter is finally written.



Starting Tuesday, researchers from four big universities and other research outfits gathered on the Carnegie-Mellon campus in Pittsburgh for the initial planning session of the “100 by 100” consortium. With a $7.5 million grant from the National Science Foundation, the group is spending the coming few years thinking about how to improve the Internet so that 100 million U.S. homes can have everyday speeds of 100 megabytes a second.



That’s more than 100 times faster than most high-speed home connections today. Files would fly across the Web almost as quickly as if they were taken off your PC’s disk drive. Imagine real-time, HDTV video links between grandparents and grandkids.



This being the Internet, in all of its free-wheeling, global splendor, the 100 by 100 consortium isn’t the only group thinking about the Net’s future. Darpa, the Pentagon agency that created the Internet in the 1970s, is also sponsoring next-generation research under a group headed by MIT’s David Clark. The two groups’ work — there are others — is seen as complementary.



Most people think that improving network speeds is a simple matter of installing faster pipes. But Prof. Hui Zhang, the Carnegie Mellon computer-science professor who heads the consortium, says even with so-called fat pipes everywhere, today’s Internet might not “scale up.”



The Internet, he explains, can’t continue to evolve with the same basic design set down a generation ago. “The Internet has been a huge success,” he says. “But the chances are that we are setting ourselves up for a great failure.”



Some of the reasons for this concern are obvious to even the most casual Web user, such as today’s chronic problems with spam, hackers and the rest.



But Prof. Zhang thinks the vastly bigger obstacle to the Brave New Web involves something more subtle: the growing complexity of the network. Much of this is unseen to average users; it’s deep in the software standard used to transmit messages — known as IP, or Internet protocol.



The professor explains the problem: The routers that serve as the Web’s traffic-control devices are so complex that only a few companies can build them. What’s more, keeping a big network running is getting harder and more expensive — “a black art,” he calls it.



Dealing with these issues means putting a number of once-solved technical issues back on the table. That’s where the evolution vs. revolution debates come in. For example, should the Internet be “connectioned” or “connectionless?” Right now, it’s the latter. All communications are tossed into the same big pipe, with routers making sure things get where they ought to.



But one school of thought says the future Internet needs to have something of the “connected” flavor of the old-fashioned telephone network, in which a direct link is established between you and the person you’re talking to.



In the world of data-communications types, things don’t get any more contentious than this.



As far as pipes, Prof. Zhang thinks that because of the 1998-2000 telecom bubble, there are enough fiber-optic lines buried in the U.S. to handle all of the backbone, “long haul” traffic of even the fastest Internet.



However, connecting up homes — the “final mile” problem — remains tricky and expensive, though new ways of using wireless communications, including reallocating some or all of the radio spectrum, could help.



Prof. Zhang acknowledges he stands on contested terrain when he says the Internet can’t continue to simply make incremental progress and expect to reach the goals of the 100 by 100 program. The computer industry is full of technologies, such as Intel’s microprocessors, that were once written off as dead ends, but which proved resilient under relentless commercial pressure.



What’s more, in this evolution vs. revolution debate, the revolutionists have another challenge. Networking companies, which weren’t around when the initial decisions about the Net were made, might oppose any technical changes, no matter how well-deserved, that threaten their market positions.



Prof. Zhang said that as revolutions go, his would be fairly staid. Most of the changes he’d want to make to the Net would be built on much of today’s system. The biggest change would involve new equipment, like switches and routers — though Prof. Zhang notes they would probably be changing under any circumstances.



Prof. Zhang is keenly aware of the PR tricks involved in getting people to move to a new technology. The current high-speed Ethernet system used by office-computer networks has little in common with the much slower Ethernet designed decades ago. It keeps the same name largely for marketing reasons, to give people a sense of continuity and easy migration.



Similarly, “whatever kind of gadgets we are going to be making in 10 years,” he says, “we will still call them ‘routers.’ ”

More here.