Why doesn't Savvis follow Savvis model?

How about giving us a brief description of what the Savvis
model is?

> I've seen several comments over the last year or so about
> how Savvis doesn't follow the "Savvis model".

As far as I understand it, the Savvis model was to set up
"PNAPs" (private NAPs) which have a DS-3 from UUNet, Sprint
and (at the time) MCI. There were something like 8 of these
across the country. The PNAPs were interconnected with their
own (probably public ATM) network, and that same network was
supplemented with public peering at the public NAPs.

At one piont, Savvis was offering one-hop (logical)
connectivity to all 8 PNAPs for customers who could do ATM,
but when I called about it, it had apparently died before it
made it off the drawing board.

The idea was that they could build a network with all of the
performance (and more) of the Big 3, from the start.

Technically, it seems to make a lot of sense. The question
always was whether it made sense financially.

Another question: if Savvis no longer follows the "Savvis
model", what model do they follow? Their Web site [1] still
talks about direct transit to each PNAP, though I noticed
that they removed a lot of the information that used to be
there (diagrams, etc). Are they moving to more centralized
peering/transit arrangements, like the rest of the

Is Savvis still considered to have a technically unusual
backbone design? How is it that they can offer the SLA [2]
they do (guaranteed average RTT anywhere on the internet <
200ms, guaranteed average packet loss < 5%, 100% uptime)
without a different network design?