Ironically, when Schaeffer began kicking about the ideas that led to Cogent, he didn't feel great urgency at first. He'd just stepped down as chairman of Pathnet Inc., a telecom company that he'd founded in 1995. From his own dabbling in the commercial real-estate market, he knew that many tenants weren't happy with the speed or the price of their Internet connections. In March 1999, he and two aides, who had joined him from Morgan Stanley and from Andersen Consulting, began researching new ways to link business customers to the Internet. Gradually, they realized that there was a huge opportunity to build something cheaper and faster.
Wanting to test his theories, Schaeffer booked a meeting with Brad Kummer, a long-time researcher at Bell Labs whom he had wanted to hire at Pathnet. As Schaeffer sketched out his ideas, Kummer didn't snicker. Instead, he declared: "Surely somebody must already be doing this."
When Schaeffer assured him that the answer was no, Kummer goaded Schaeffer to act fast. Schaeffer did -- and recruited Kummer to be Cogent's chief technology officer. Since last February, Cogent's workforce has been climbing by nearly an employee a day. One of the duo's first priorities: sketching out the design of Cogent's backbone, starting with the fiber-optic cable.
On its own, Cogent couldn't hope to install roughly 12,000 miles of fiber in a matter of months. But major long-distance carriers had been overhauling their own networks, and many had put in so much capacity that they had plenty of strands of unused, or "dark," fiber that Cogent could buy. A single strand of fiber, no thicker than a piece of human hair, can carry more than a trillion bits of data per second.
Schaeffer and his team quickly settled on Williams Communications, based in Tulsa, Oklahoma, as the major source of their fiber network. Williams, a onetime gas pipeline company, had shrewdly used its rights of way to diversify into telecommunications. At times, Williams had strung fiber-optic cable inside its old pipelines. In other regions, Williams needed dozens of backhoes to dig afresh. Its network didn't provide comprehensive rural coverage, but that didn't matter to Cogent. For linking to big downtown office towers across the United States, Williams's network was perfect.
By early September of this year, Cogent had bought or had leased strands of fiber that covered much of the United States in a huge, sideways figure eight. Starting from a midpoint in Kansas City, Missouri, one giant loop reaches east to connect such cities as Dallas, Miami, Atlanta, New York, Boston, and Chicago. Another loop stretches west to connect Denver, Phoenix, Los Angeles, San Francisco, and every major city between.
In order to offer such high-speed service, Cogent needed to own that Internet backbone itself. Typically, coast-to-coast Internet traffic takes about 13 "hops" from one network to another, searching for routing paths that will get data to its destination. Each hop creates a slight delay in moving the data along. In aggregate, those hops can slow transmission to a crawl. Cogent, by contrast, aims to eliminate almost all of those time-wasting hops -- thereby speeding up users' connections -- by keeping data on its own network for nearly the entire journey.
But Cogent's nationwide Internet backbone would be splendidly useless if it stopped at the outskirts of each big city, like an interstate highway system without off-ramps to local roads. Thus, Schaeffer's company has bought 5,000 miles of short-haul urban fiber from Metromedia Fiber Network Inc. of White Plains, New York. Cogent also is using Cisco's optical routers to manage traffic, so messages get to the right place.
A few years ago, it would have taken $3 billion or more to install enough equipment to keep Internet traffic running smoothly nationwide. That's why only a handful of national Internet backbones have been built to date -- and those were done chiefly by long-distance phone carriers that could afford the massive outlays involved. Instead, traffic is often handled by a bucket brigade of regional networks, pitching in to move data from origin to destination.
But now, with routers and other equipment getting faster and more powerful, even a startup can think of covering the entire United States on its own. Schaeffer also decided to keep costs down by building a network that could handle only Net traffic, without being able to carry voice calls as well.
Cogent's financial forecasts suggested that the new system could break even if it were used by seven to nine tenants each in about 800 office towers across the United States, with each user paying $1,000 a month for its Internet connection. By contrast, corporate Internet users typically are paying about $1,400 a month for the telecom industry's current benchmark for high-speed service: a T-1 connection that functions at slightly more than one megabit a second -- barely one-hundredth of the Internet speed that Cogent promises.
Recent Comments | 1 Total
February 24, 2009 at 3:09pm by Eli Shapiro
Well many years later and this company is still an unknown in the market, with US high speed internet ranking very badly among the offerings available worldwide. I wonder what happened with this fiber project that derailed its success? If they're still in existence, though, there's a huge amount of stimulus money being set aside for creating a new backbone, which it seems like they already have experience doing...