Deep Ignorance on the Optical Front
by Steve Steinke, Editor-in-Chief
Network Magazine
12/05/00, 12:47 p.m. ET
We've heard about Moore's Law for 30-some years or all our lives, whichever duration is longer, and sure enough, those computers just get faster and more powerful all the time. Optical fiber came out of the labs in the 1970s, and sure enough, a strand of fiber “no wider than a human hair!” can carry 16,000 times as much data now as it could 20 years ago. I'm in this field because I appreciate technological changes, but I long ago began to take for granted exponential increases in CPU capability, RAM capacity, disk drive space, and network backbone throughput. I don't think that having adapted to these kinds of continuous change makes me some kind of grizzled, leather-skinned cynic.
The authors of an October 9, 2000 Business Week article claim that 160-wavelength Dense Wavelength Division Multiplexing (DWDM) “has torpedoed the cost of capacity for a phone company by more than 99 percent.” It's true that a single fiber suddenly carrying 160 channels where it previously only carried one is 99 percent less costly per bit. It's true that the extra capacity doesn't require new trenches and rights of way. However, the multiplexers, switch and router ports, lasers, amplifiers, and regenerators that are required for each wavelength aren't negligible expenses. Furthermore, 160 times the traffic may well mean 160 times as much customer service, as well as some sizeable multiple of the costs of billing customers on a single channel.
The reason DWDM is almost exclusively a long distance technology today is because the ratio of the cost of fiber and a trench to the overall cost of the service is high-it maximizes the advantage of the technology. However, in access networks, and especially in residential and small business settings, the applicability of DWDM is much less apparent. In many respects, the long distance network can be compared to the railroad network, and the ability to carry multiple wavelengths is comparable to developing new locomotives that can pull, say, 160 freight cars. It's true that each train will carry more freight with the new locomotives, but the mechanisms for loading each car and hooking them up to the train will be proportional to the number of cars-they won't have the same economy of scale as the newly possible long train itself.
Furthermore, the existence of 160-car trains has next to no relevance to someone who wants a pizza, or even a pickup-truck load of sheetrock, delivered to his house. Similarly, DWDM is a great technology for massive, aggregated long-haul traffic, but it's not particularly attractive for the access network.
What makes me the maddest when nonspecialists write about network technology is the cheerleading element. Sometimes the best thing a technology writer can do is refrain from describing the subject as revolutionary, a paradigm shift portending inevitable doom for the dinosaurs of the moment. The prime culprit here is the former Reagan speechwriter, George Gilder. Unfortunately for technology, and ultimately for the packs of investors who have come to swear by him, Gilder is a fabulous writer. Every engineer he interviews sounds like a genius; every startup CEO is a visionary. The exponential growth of fiber capacity has been flogged by Gilder for almost ten years now as the greatest engine of economic transformation in history. And the “supply-side” or Field of Dreams rationale, “If you build it, they will come,” gets trundled out whenever anyone dares to ask, “Do I really need an OC-3 link to my house, or my small business, or even to my factory?”
I'm convinced that if service providers such as Cogent Communications can continue to exist based on selling 100Mbit/sec connections for $1,000 per month, then access fiber networks will never extend beyond a handful of dense metro areas, because the payback period will be too long to justify expansion. Widespread fiber-to-the-customer-premises deployment will only happen if someone can make money building it; if the cost of fiber-based bandwidth approaches zero with advancing time, then the build-out will stop. The worst enemy of an expanding role for high-capacity fiber optic technology is the dropping cost of bandwidth-ironically driven by the real advantages of fiber and kicked into overdrive by journalistic cheerleading.
networkmagazine.com |
