Under Construction
The federal government continues to invest millions of dollars in its part of the advanced Internet projects; over the past three years, it has spent US$272 million - which officials regard as seed money. Much of the money is awarded in the form of competitive grants for either a specific research project or to assist universities in paying for the connection to an ultra-high-speed backbone. The commercial cost for an OC-3 (155-megabit-per-second) connection to one of the backbones is US$21,600 per month.
The government also spends hundreds of millions of dollars on special high-speed network projects for some of the most demanding agency uses: the DOD, which seeks the most advanced battlefield intelligence; the DOE, for modeling atomic blasts and being able to maintain the nation's nuclear weapons stockpile without conducting actual bomb explosions; and NASA, for the complexities of space missions and communications.
But the government is, in fiscal terms, a junior partner in most of the efforts. Not only do universities commit resources to advanced Internet research, but also corporate co-sponsors provide money and in-kind services worth more than the government's contribution. Companies such as Cisco Systems, IBM, Microsoft, Nortel Networks and Qwest Communications International have made substantial investments in partnerships with universities.
Qwest, for example, has made a five-year commitment to the UCAID to lend the consortium an enormous fiber-optic network. With 10,000 miles of fiber capable of OC-48 (2.5-gigabit-per-second) speeds, the network would cost US$500 million to lease commercially, says Greg Cook, vice president of Internet services at Qwest.
"It's not a lot of money, and it's highly leveraged," Howe says. Advanced Internet research in the U.S. actually comes in two flavors: Internet2, where the government's direct role ended in 1999, and the NGI, which is largely government sponsored.
Internet2 was started in late 1996 by a consortium of 34 universities that were concerned about creating a test bed for large-scale networking technology and applications outside of the commercial Internet. Federal dollars helped build what are now two distinct fiber-optic backbones in the U.S.: one called VBNS, now operated by WorldCom; and the other known as Abilene, managed by Indiana University and operated by Cisco, Nortel and Qwest. There are nearly 200 universities and research labs, including government labs, connected to the two backbones, says Greg Wood, a spokesman at the UCAID. There are about 70 corporate partners as well.
The NGI is a federal research project involving most of the agencies that have advanced networks. It is overseen by the National Coordinating Office for Information Technology Research and Development and by the President's Information Technology Advisory Committee.
In effect, much of the work is interconnected in a way that former Novell Chief Executive Ray Noorda would have called "co-opetition": While the different projects and networks compete to develop technology and software with potential uses in both research and commerce, there is a relatively open exchange of information.
"Everything we do is available to people, and we don't have any sort of intellectual property constraints or trade secrets," Van Houwelin says. But he acknowledges that companies that become partners have a leg up on their competitors because "it is the case in the Internet world that people who do it first get the advantage."
Indeed, the projects have exposed some problems that aren't at all technical in nature.
Take an area such as quality of service - in essence, being able to guarantee that a particular time-dependent application's data, such as a videoconference, will receive priority on the network, so the bits will arrive in time to be reassembled into a usable image. Researchers say the chief problem uncovered with QOS isn't technical - indeed, such guarantees are offered by some commercial networks today - but economic. Nobody is sure what business model will work to allow QOS guarantees to span multiple Internet service providers and ensure that everyone is appropriately compensated.
The advanced projects, though, have brought corporate partners together to begin working out such economic details as well as the technical ones, the partners says.
Most of the researchers say there are a few significant areas in which improvements over the commodity Internet are already being realized in the advanced networks:
Multicasting: Listening to streaming audio or watching an online video on today's Internet can be a strain for many users. It can also be a strain on the provider. Each user who wants to view the program must have an individual video stream sent to the target computer. If enough people want their own streams, the capacity of the server to deliver them can be quickly exhausted.
Advanced networking researchers are working on the technology needed to solve the problem. The new approach to multicasting would permit the source server to send out a single data stream that would be "mirrored" by other servers. Any number of users would be able to tap into the stream along the way.
Multicasting is already being used on some campuses to deliver lectures and other educational programming to students who aren't in the classroom. For commercial use, researchers are working on adding security components to keep private transmissions closed to all but authorized users.
Multicasting could be one of the first aspects of advanced network research to become a mainstream public phenomenon: It's one of the cornerstones of future entertainment products that could include three-dimensional and virtual reality components.
Optical Switching: The basis of all advanced networks today is a reliance on fiber-optic connections because of their speed and capacity. But that also means researchers have to figure out how to build optical switches that can work faster and increase capacity on a given piece of fiber.
One of the most important advanced network projects in this area, Howe says, is the development of Wavelength Division Multiplexing. That technology sends many different sets of data along an optical fiber by assigning each one to a particular wavelength of light. It is similar to using different frequencies on a copper wire to handle various types of data, except that fiber optics are far faster and more versatile.
Distributed Applications: The original Internet and its successor, the NSFnet, allowed researchers to make use of the power of distant supercomputers. The advanced networks make it feasible to think about doing the same kind of work in the future without needing a single, expensive machine. By distributing a computing problem around the network, legions of smaller computers can devote unused processing time to solving the problem.
The idea isn't new. Hackers have used it to break encryption schemes; astronomers use it to check radio telescope data for signs of extraterrestrial intelligence; and a new dot-com relies on the scheme to help its corporate partners search for an AIDS vaccine.
But those and similar efforts today are "embarrassingly parallel problems," ones that can trivially be broken into small parts that can then be worked on in isolation, says Todd Needham, manager of the research programs group at Microsoft. For example, the SETI@home project, which searches for patterns in radio astronomy data, simply divides its massive data into chunks of a few hundred kilobytes. Each PC involved in the project takes a chunk at a time, runs the data through some equations and passes the results back to the astronomers.
Work being done on Internet2 involves how to put intelligence into the network itself to manage problems where things simply can't be divided into neat chunks. The network will be able to quickly rearrange itself, farm a part of the problem out to an available computer and coordinate the dynamics of the process. That would let researchers perform "pretty close to the full set of supercomputer applications today" on the network, Needham says.
Distance Instrument and Process Control: One of the most intriguing uses of the advanced backbones to researchers at both universities and government labs is the potential to control complex - and usually exceedingly expensive - instruments from long distances.
For example, take particle accelerators, enormous machines used by physicists to break matter down into its most fundamental bits and study them for clues to the basic rules of the universe. A new device is being constructed for this purpose at CERN, the Swiss facility where Tim Berners-Lee invented the World Wide Web. With advanced backbones and the ability to move the huge data that results to remote machines, CERN's accelerator and similar facilities can be used by scientists all over the world without forcing them to travel to the device itself.
"Scientists believe this is going to revolutionize physics," Strawn says.
Similarly, astronomers will be able to coordinate their use of telescopes around the world from their desktops, and computer scientists and other researchers will be able to command the fastest supercomputers from remote sites, or even in collaboration with other researchers at different institutions.
While that may not sound like a direct benefit to the public at large, experts say the same technology and techniques can be applied to remote medicine, perhaps at some time allowing re-mote surgery. That would bring the benefits of the best medical knowledge available to nearly any patient.
"Researchers and medical practitioners can have direct physical access to, say, an electron microscope located physically on the other side of the country," says Gordon Wishom, chief information officer at Georgia Institute of Technology. "There are a number of applications in that space that have been very important to researchers that can have a direct impact upon people's lives."
Security: The original Internet was never designed with security in mind - possibly because its users were all supposed to be cooperative, possibly because it was open only to government-approved users and possibly because nobody imagined hackers.
Getting It Right
With security as one of the key flaws in the commodity internet, researchers are busily using the advanced networks to figure out how to make stronger - and more modern - types of security ubiquitous.
"If you look at the basic concepts of public key cryptography, the most important conceptual breakthroughs happened over 20 years ago," says AT&T's Bellovin, one of the world's top computer security experts.
His work involves figuring out ways to provide encryption and authentication for sensitive traffic, and the creation of a network design in which every transmission is encoded and all Web sites and users hold a verifiable certificate proving they are who they say they are.
The advanced network research has even spawned more ambitious projects. Cerf, for example, heads a group based at NASA's Jet Propulsion Laboratory in Pasadena, Calif., that is examining how to make the Net work on an interplanetary scale, so that spacecraft and, perhaps, humans on future missions will have a standardized and reliable method of communication with Mother Earth. The first satellite equipped with a prototype of the necessary software is slated to be launched this month, Cerf says.
And to critics who worry that things are making their way into the commercial sphere a bit too slowly, at least one researcher warns that you might want to let them get it done right instead of quickly.
"I hope it will be a couple of years before I'm operated on remotely," says Craig Labovitz, a research scientist at Microsoft.
