[nsp] sup720-3bxl experiences?

[Rajeshbansal9 at aol.com]

We did some testing using  an agilent and got 500k of bgp routes before the 
sup7203bxl took a nose drive. Cisco advertises 1 million routes. Anyone got 
anything higher than 500k routes?

thanks

raj


In a message dated 3/6/2004 6:36:58 AM Eastern Standard Time, he at uninett.no 
writes:
The common rule of thumb established in the early 90s saying that you
need at least BW x delay amount of buffering on ports which form a
bottleneck in order to support high-performance TCP over the path
still holds true today, as far as I know.  It is possible that active
queue management in the form of WRED can modify this somewhat.  Note
that the delay here is between the endpoints of the TCP connections
you want to support, not the local delay over the bottleneck.  So, in
ISP backbone I would say that you would do well to stick to this rule
-- it's plain conservative engineering.

In addition to this I've been told that there is no internal flow-
control in the 6509 / 7600 switches, so that if an output queue on a
particular port runs full, you will in effect not have any other
buffering to help you out to prevent packet loss.

Regards,