Fiber Famine
by Bret Swanson
The fiber glut, as much as anything, is the conventional symbol of our economic downturn. To most it represents over-capacity, excess—greed. Telecommunications industry executives and technologists laugh at the glut theory as a simplistic, uninformed cliché. But until now no one has offered numbers that show just how wrong the herd of doubting journalists and Wall Street analysts has been.
Maybe more than any other company, fiber network operator Williams Communications (WCG) has taken up the charge this year of refuting industry myths and pointing to the long-term promise of the Internet. For months CEO Howard Janzen has explained how Williams, along with Level 3, Metromedia, Broadwing and others, bury 20 years worth of fiber in order to economize on the labor-intensive, time-consuming tasks of trenching and pulling fiber through thousands of kilometers of plastic conduit. The lasers, receivers, amplifiers, and switches that transmit the bits, moreover, cost some 10 to 20 times more than installing the fiber itself. Dark glass cannot, therefore, be considered "over-capacity."
Evidently, investors have been unimpressed with the cogent explanation, feeling in their guts that five next generation fiber optic networks is too many. But here's where the new numbers come in.
On December 6 and 7 Williams held a two-day educational seminar for reporters and investors at its spectacular new headquarters building in Tulsa, Oklahoma. Broadband University's most striking presentation was offered by TeleChoice's Russ McGuire, who showed that within four years the fiber glut may become a fiber famine.
Contrary to the New York Times and Wall Street Journal's famous assertion that just 2.6 percent of the nation's fiber is in use, McGuire's statistics compiled from 29 U.S. carriers show that on the 22 most popular inter-city, long-haul routes, 3,400 out of 15,177 total fibers are "lit." That's 22 percent of the installed base of silica strands. Fourteen of the 22 routes are running at 70 percent of capacity or more. More stunning than the current figures, however, are McGuire's projections.
If by 2005 just 7 percent of U.S. businesses adopt Gigabit Ethernet connections to the Internet, the Telechoice MADCAP model shows that all the capacity on every fiber on each of the 22 most popular routes will be used up. In other words, we achieve total long-haul fiber exhaust—no small feat considering Williams, Level 3, and the other new players have so far lit between 2 and 6 fibers out of 100, depending on the route. Five percent Gig E penetration exhausts almost all the fiber currently available, and 2 percent penetration exhausts about half the routes. The success of Ethernet as the connectivity solution of choice is far and away the most important factor in "filling the pipes," more than new killer apps or anything else.
The next generation carriers have all built extra conduits, some more than others, through which new and better fiber can always be pulled. Thus fiber exhaust is not a problem but an unmistakable signal of success. After the substantial capital costs of building networks have been paid, selling massless photons is a very high margin business.
McGuire's MADCAP makes fairly moderate assumptions on how fast wavelength division multiplexing (WDM) will advance. If you believe most of the new capacity between now and 2005 will be of the 10-gigabit (OC-192) variety, MADCAP assumes the bulk of new bandwidth will be deployed using 160-channel and 320-channel systems. It is likely by 2005 we will see WDM systems capable of carrying several thousand channels on a single fiber, not considered by the MADCAP model, so more aggressive WDM may push back the fiber exhaust doomsday a year or two. (We should remember, however, that as channel counts proliferate wildly bit rates may actually start falling. A 320-channel, 10-gigabit system is 3.2 Tbps. With today's usable fiber window of about 90 terahertz, the most capacious system we can envision in the near future is therefore about 9 Tbps, just 3 times MADCAP's maximum.)
McGuire's findings shift the attention to Ethernet connectivity providers Yipes, Cogent, and Telseon, and even to Narad Networks, transformer of coaxial cable plants. The success of these companies, getting fiber (and coax) into buildings and productizing their bandwidth capabilities, which are some 100 or 1000 times better than today's T-1 offerings, will determine the future of the long-haul carriers.
gilder.com |
