On Thursday 21 August 2008 00:47:35 Everton da Silva Marques 
wrote:

> 1. Use RR to scale BGP; deploy RRs in pair for the
> redundancy Keep RRs out of the forwarding paths and
> disable CEF (saves memory)

Curious to actually know by what % memory is saved when CEF 
is disabled on a fairly modern IOS-based platform today, 
particularly, software routers (which make affordable route 
reflectors).

We use 7201's for route reflectors (IPv4, VPNv4, and with 
any luck, l2vpn AFI in the future), with tens of peers 
configured, about 5 of whom send full tables, while the 
rest send a couple of hundred to a few thousand routes.

Configured with 2GB of DRAM, 1.78GB is available after boot 
(thanks to 12.2(33)SRC1), and 322MB is used to hold the 
learned routes (as at today).

1.47GB is left for future use.

CEF is enabled, CPU lays low at 1% to 2%. Overload bit is 
set on IS-IS to keep them from being transit routers.

With these values, we're not inclined to disable CEF; but 
would be interested in hearing your experiences with this 
practice, and on what platforms.

Cheers,

Mark.

Attachment: signature.asc
Description: This is a digitally signed message part.

_______________________________________________
cisco-nsp mailing list  [email protected]
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

Reply via email to