Its 5 OClock, and Its Much Too Crowded

Scientists have been studying road traffic patterns for years, yielding absolutely no discernible results.

Scientists have been studying road traffic patterns for years, yielding absolutely no discernible results. Now, according to the San Francisco Chronicle, scientists are working on a "unified" theory of traffic, an idea I fully support, even if the randomness of the world makes such theories implausible.

Alas, theres no unified theory of Internet traffic, but it appears that a lot of intellectual horsepower is going into studying the issue. So far, most of the companies that want to increase performance simply want to slap bigger pipes onto the Internet. Nortel and JDS Uniphase wish everyone had optical pipes running everywhere, the equivalent of adding a 42-lane highway across the country. What happens when theres a traffic circle in the middle is not really their problem.

Other companies want to put in private peering arrangements so that traffic mostly flows over their networks. AboveNet and Exodus, for example, have set up peers in which all the traffic from one AboveNet co-location space is whisked to another AboveNet co-location space in record time. This is the equivalent of the toll road.

Then there are the companies that want to cache data so you dont have to travel at all. Akamai and Inktomi, for example, have made big names for themselves with this technology.

One company, however, is studying the real bottlenecks—those that appear at the exchange points themselves. InterNAP is the only company I know of that has paid arrangements with each carrier—UUNet and Sprint, for example. InterNAP has private network access points to ensure customers are always on the cost-optimized route.

Unfortunately, InterNAP has to figure out how to measure its value, and to do so it must measure Internet performance. And it doesnt want to rely on Keynote Systems, a company that creates the most unique grimaces on InterNAP officials faces. Some of InterNAPs key customers, including, are on Keynotes public performance metric and never appear in the top-five fastest positions.

I sense a war brewing. I also sense that InterNAP will broaden its network offerings. If it doesnt, I suspect that InterNAP might be unifying with another company—whether it wants to or not.