a little thought on exchanging traffic

Anyone thought about eliminating large physical exchange points and replacing
them with a more distributed architecture?

Yes, lots of folks have given it lots of thought.

Multiple data centers interconnected over ATM in a single metro area run by
indepdenant entities who are free to provide any level of service or value
add they wish.

The fundamental problem is there are no magic pixie dust in this business.
Sure, some people like to put out press releases saying how they've solved
all the worlds problems using the Magic Frambulator. But what they've usually
done is ignored half the problem.

The Sanfrancisco NAP is actually a distributed architecture of several ATM
switches around the bay area. And although the SF-NAP is handling far
less traffic than MAE-East on a daily basis, it too has suffered from the
problems of congested trunk lines between the switches. There have
been other proposals too.

Most of the 'solutions' I've seen have had either far fewer number of
interconnects, or far less total cross-connect bandwidth. It seems to
be very difficult to get both. When they get close to either the number
of connections, or the amount of traffic, they seem to run into the same

Hey, I saw one of those new Magic ones last week at "Frambulators R Us".

Way too expensive though, we've had to stick to good old fashioned Clockwork
Frambulators (electric ones still being unreliable, and steam-powered ones
outlawed under Californian emissions legislation).



Ohmygod! We're out of magic pixie dust!?! The Internet will collapse in 30
minutes!!! How could the Internic let this happen?

I'd like to propose an RFC for conserving pixie dust. Maybe a non-profit
arm of the Internic to allocate magic pixie dust on a fair and equal