In Global Contest to Build Networks, Does the Race Go to the Swiftest? New Canadian network has world's fastest backbone, but few applications to use on it By JEFFREY R. YOUNG Quebec Canadian universities now have the fastest research-network backbone in the world, officials here said during the annual meeting last month of Canada's top network gurus. At the meeting, the usual geek speak of zeros and ones was accented by a sense of national pride. "We are first, and it feels good," said Alan Greenberg, director of computing and telecommunications at McGill University, during a conference session. Such pride is well-founded. The new Canadian network, known as CA*Net3, can transmit the contents of the U.S. Library of Congress in one second, according to the network's architects. It would take a minute for the U.S. high-speed network, created by the Internet2 consortium, to do the same job. In the United States, where the Internet was invented, network officials point out that the Internet2 project supports more research into new network applications than does its Canadian counterpart, and that such applications are the keys to spurring further revolutions in on-line communications. "What we would be envious of is if they had wondrous applications that they're making use of," said Ira Fuchs, vice-president for computing and information technology at Princeton University, who has been involved with Internet2 since its inception three years ago. The competition to build the fastest new network and tools -- call it the cyberspace race -- is heating up, and universities are leading the field. The stakes are immense, at least according to the rhetoric. In the information age, it is said, the country that plants the biggest flag on the network frontier will wield the most power and create the most jobs. "A lot of countries around the world would dearly like to make us second fiddle," says David J. Farber, a professor of telecommunications at the University of Pennsylvania, who is also a long-time leader in the Internet2 effort. "The technology isn't letting us rest," he adds. "Big countries and big companies can go from very big to very small very quickly." Some network experts say they aren't building fast networks to beat other countries -- they're doing it to satisfy the needs of researchers. And they say that any rivalry is friendly, at least among those building the networks. In fact, the folks building Internet2, CA*Net3, and advanced networks in other countries work together closely, share information, and interconnect -- or "peer" -- their new networks to make their tools accessible to as many scholars and researchers as possible. Computing officials say they have a lot to learn from one another's network projects. Canada has a tradition of leadership in fiber-optic networks, and the country is home to several technology companies that have done significant research in the field. By many accounts, a major force behind the development of CA*Net3 was Bill St Arnaud, the senior director of advanced networks for CANARIE, the Canadian Network for the Advancement of Research, Industry, and Education. The group is a consortium of universities, businesses, and government agencies that is the Canadian equivalent of the Internet2 group in the United States. Mr. St Arnaud believes that the technology behind CA*Net3 represents a revolution in design that will soon make networks faster, and also much cheaper, for universities and consumers. As he explains it, the backbone of today's Internet is made up of several layers of network technology that serve various functions in directing the flow of data. Each layer helps guarantee reliability, but also increases the cost and decreases the speed of the network. The multilayered networks were originally designed to carry telephone traffic, not data, he says. The trend in high-speed data networks over the last few years has been to remove such layers, allowing more room in the data pipe for "payload," such as e-mail messages, music files, or video streams. But to keep such a stripped-down network stable, engineers rely more heavily on "intelligence at the edge" of the network, such as routers and other network hardware that is closer to the end user. And it helps that "packets" of Internet data are encased in digital envelopes called I.P., or Internet Protocol, that include routing information. Computers and routers on the network can read I.P. information and reroute data when trouble arises. One reason that Internet2's latest backbone is faster than the current Internet's is that it removes a network layer called Asynchronous Transfer Mode, or A.T.M., which older systems rely on to guarantee that information arrives with all its parts intact. Dispensing with the A.T.M. layer allows more data to travel over the network at the same time. For CA*Net3, Mr. St Arnaud proposed embracing an approach that went a step further: removing the Synchronous Optical Network -- or SONET -- layer, which experts describe as a "self-healing architecture." SONET can redirect packets around a problem area, such as when a backhoe severs a cable by mistake. The Canadian network also uses another new approach that vastly increases the performance of the fiber-optic strands on which the data travel. Engineers have divided light beams into several wavelengths, or individual color bands, and can send different data signals on each wavelength. Network experts are still experimenting with the technique, which is called Wave Division Multiplexing, or W.D.M. The new network was a hard sell, says Mr. St Arnaud, especially because some of the businesses in the CANARIE consortium believed that cutting out the SONET layer would make the network so unstable that it would never be marketable. To complicate matters, CANARIE was already building CA*Net ii, a high-speed network with A.T.M. and SONET layers. In the end, CANARIE decided to take the plunge, and it secured $55-million (Canadian) from the Canadian government to support the effort. Then officials started from scratch, building a fiber-optic network designed to carry only research data. The new network now spans the country, from British Columbia to Nova Scotia. Initial tests seem positive, Mr. St Arnaud says. "We demonstrated to them that it does work. It works very well, and it works as well as, if not better than, existing SONET services." One of the biggest benefits of eliminating network layers, he says, is that the new network's simplified design makes it cheaper to build. He says CA*Net3 is "two orders of magnitude cheaper" than a comparable network built with older technology. Mr. St Arnaud sums up the project with a single French phrase about the benefits of transmitting data directly over fiber-optic cables: "La vitre, c'est la liberte." Glass is freedom. U.S. network officials also considered using the strategy adopted by the Canadians, but decided against it for now, says Guy T. Almes, who is the chief engineer for Internet2. One reason was that major companies Internet2 works with, such as Quest and Nortel, were eager to test their latest SONET equipment. The companies offered U.S. universities a cut-rate price on access to SONET equipment that was "simply too attractive" to pass up, says Mr. Almes. And, Mr. Almes adds, "to have gone down the W.D.M. road initially would have added complexity and delay to what was already a very ambitious network design." Besides, Mr. Almes says, Internet2 is more interested in developing new network applications than in building the world's fastest backbone. The speeds available on the Internet2 backbone, Abilene, are plenty fast to support new applications that could lead to revolutionary ways of communicating, he says. Canadian network experts agree that in the realm of applications, Internet2 is in the lead. At a discussion at last month's network conference here, participants complained that too few researchers had come forward to experiment with the CA*Net3 network. Said one network administrator, "We're so busy just putting the pipes together, we have no time to go into the university and actually talk to people who use the stuff." An Internet-services provider agreed. "I'm frustrated, because Canada is leading the world in networking, but the applications are lagging," said Peter Sutcliffe, president of Meissa Inc., in an interview. "We're going to have the best network, but we're going to be delivering everyone else's content." Another conference participant said Canada needed a kind of "man on the Moon" goal -- a concrete project that could be done only with a high-speed network. "We need someone who's willing to stand up and say, 'We want to put this application on the map for Canada,'" said Laurie Mersereau, a customer-systems engineer for NBTel, a company in the CANARIE consortium. Andrew Bjerring, president and chief executive officer of CANARIE, said that encouraging the development of applications was now a top priority. At the conference, he announced a three-year, $8-million program to support work on advanced network applications and services. Even in the United States, researchers say they still haven't found the "killer application" for Internet2. Applications under development include videoconferencing, controlling scientific instruments from a distance, and research collaborations in virtual-reality environments (The Chronicle, August 8, 1997). "It's going to be a graduate student out in some distant place that's going to make a dramatic change," predicts Mr. Farber, of the University of Pennsylvania. It may very well be a graduate student outside the United States. Researchers in the Netherlands are building a superfast network called SURFnet. Scandinavians are building one called NORDUnet. And Japan and other nations are developing the Asia-Pacific Advanced Network. So the race continues -- as does the rhetoric. Rita R. Colwell, the director of the National Science Foundation, said in a speech early this year that it was "America's information destiny" to lead the world in networking. She didn't mention that other countries were at its heels. http://chronicle.com Section: Information Technology Page: A21