NSF Brings New Order to Internet

P> A class structure is officially on its way to the Internet.

The NSF said it will remain neutral on what methods are used by grant award winners to eliminate the Internet bottlenecks. However, Mark Luker, who took over as director of the NSFNET program in December and will oversee the "Connections to the Internet" grant program, said prioritizing use of the Internet is a likely solution.

The problem is clear to even the most casual user: The Internet is overburdened. At the beginning of 1996, the Internet included more than 100,000 networks that interconnect millions of computers and tens of millions of people worldwide, the NSF said.

Longtime Internet users, especially in the scientific community, are finding that supercomputer projects and other scientific programs are actually being destroyed by the on-line popularity. Some projects just creep along, others produce unreliable results, and some simply don't work. Examples of such projects include interconnecting supercomputers, remote control of instruments such as telescopes and microscopes, and control of medical procedures through a network.

"Today on the Internet you are not able to get the service you need even if you can pay for it," said Luker. "There needs to be some correlation between supply and demand."

Luker compares the idea of prioritizing Internet traffic to the U.S. Postal Service, which offers options such as overnight, first-class or third-class delivery. The difference, however, is that although the customer chooses the method of sending a letter, the government will regulate who gets e-mail priority. Companies might have to fight to get to the top of the priority list, and the result will be controversial.

The way the Internet works today is that all Internet traffic is created equal. A prioritization structure would favor government and scientific endeavors, as well as large businesses over small businesses and individuals.

By programming switches and routers to allow certain classes of information to flow before others, a technological equivalent of a high occupancy vehicle in a carpool lane would be created. Besides type of user, priority might also be determined by use.

For example, video takes up more time, and therefore, might be relegated behind e-mail, or the user may have only a certain number of minutes for a certain task. It's still anyone's guess as to how the class structure will shake out. The idea has been bandied about in the industry as a solution to the overburdened Internet, but many are surprised the NSF has taken the lead role.

"Class of service has been known to be a requirement for quite a while," said Michael Roberts, vice president of Educom, a Washington, D.C.-based infotech association for higher education. "It's necessary for broadband Internet."

Brian Muys, a spokesman for Internet access provider PSINet Inc., Herndon, Va., agreed that a priority structure looks like the only solution.

In fact, PSINet and other companies already use their own class structure. At PSINet, Fortune 100 companies get top treatment over everyone else, Muys said.

However, he said he hopes the NSF will give industry some power in the new system. "The whole raison d'tre for the Internet was to take it out of the government's hands," Muys said. "You hear so little about the NSF anymore that it looks like it's gone full circle." Luker said the new program is an update of a previous connection project that was published in 1990.

However, the grant system is set up primarily to encourage educational and research institutions -- not commercial Internet companies.

Luker said he hopes to give the first award by late next fall. Deadlines will be twice a year, Jan. 31 and July 31. Any way to help direct "meritorious" traffic more quickly through the Internet will be considered. "We are not really taking a position on how to do this," Luker said. "The technology is there to accomplish this in a number of different ways."

There are three categories for the two-year grants: A grant of $15,000 for innovative technologies for Internet access for K-12 schools; a $20,000 grant for connections for higher education institutions; and a $350,000 grant for high-performance connections for research and education institutions.

The grants will fund prototype efforts that the NSF will consider in trying to bring order to all Internet traffic.

Some might use the NSF's very high-speed Backbone Network System, which connects the NSF supercomputing centers.

Luker, who is on a two-year appointment to the NSF from the University of Wisconsin-Madison, said he hopes to make about 20 awards a year.

The grant program will go on for several years to build a permanent infrastructure. "We're going to be pushing the Internet connection envelope," Luker said.