The following text is copyright 1996 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

Restrictions in the unlimited

For quite some time it has been common wisdom that there is essentially unlimited bandwidth available for the Internet and for other telecommunications applications in the fiber optic infrastructure "in the ground." It is not clear where that idea came from. It could be because people think that because fibers can be run at high bandwidth and because there is a lot of fiber already installed there must be more than the current and future applications can fill. But one thing for sure, the idea did not come from the folks that are trying to find fiber to expand their backbones to meet the explosive growth in Internet traffic. Many of them are quite worried about where the bandwidth will come from in the next few years. The existing fiber plant is becoming exhausted and new fiber builds take a lot of planning, time and money. On top of it all there may be a problem getting the fiber itself.

The New York Times reported on November 4th that the world-wide demand for fiber used in telecommunications will reach16.25 million miles of fiber this year, a market worth about $6 billion. But production has been running behind demand and the short fall could be more than 1.25 million fiber-miles this year. The demand has been fueled by the global telecommunications deregulation movement. In addition the advent of competition from direct broadcast satellites is forcing U.S. cable TV companies to undertake large scale infrastructure upgrades.

The vendors are trying all sorts of things to gain more bandwidth from increasing the data rate to wave division multiplexing (WDM). Note that capacity and bandwidth are not the same thing. Technologies such as WDM are a way to provide more capacity by making more wires, not higher bandwidth by making faster wires. Using WDM to put 4 OC48 (2.4Gbps) circuits on a single fiber does not make one OC192 (9.6Gbps) circuit. Instead, it provides 4 circuits that must be independently terminated in switch or router ports.

But whatever tricks are played, when the bandwidth demand is increasing as rapidly as it is (doubling every few months for the Internet backbones for example) it is a bit hard to keep up with.

This is not another death of the Internet is imminent column. These problems are real but not immediate. Instead I'd like to talk about another topic.

One of the features of today's Internet is that there is only one product being sold: IP connectivity. It matters little if the data being transferred is a spam message or an invoice for $3 million, if its real-time Internet telephone or a piece of email where it would be OK to take an extra 5 min. to deliver. Undifferentiated products of this type leave little room for ISPs to compete other than in price. If (or better, when) the network infrastructure can support multiple levels of quality of service, the ISPs can start to compete on a basis of quality of data delivery not just cost or how fast they answer the phone. It also may provide a way for the ISPs to moderate the growth in their bandwidth requirements enough so that the production of fiber and the installation of new facilities has a chance to catch up.

disclaimer: Undifferentiated is not a term that many people apply to Harvard -- in any case, the above worries are mine.