[4757] in North American Network Operators' Group
Re: Advice on dealing with Sprint (fwd)
daemon@ATHENA.MIT.EDU (Alan Hannan)
Fri Sep 27 18:17:54 1996
From: Alan Hannan <alan@anka.mindvision.com>
To: nathan@netrail.net (Nathan Stratton)
Date: Fri, 27 Sep 1996 17:13:03 -0500 (CDT)
Cc: markd@cais.cais.com, cosmo@olywa.net, rob@rjl.com, nanog@merit.edu
In-Reply-To: <Pine.LNX.3.95.960927175122.14285C-100000@netrail.net> from "Nathan Stratton" at Sep 27, 96 05:54:11 pm
Hi Nathan,
> > We do this frequently and with a 2501 or a 2514. It must be a Cisco router
> > and sometimes will need more memory. There is NO additional charge for those
>
> Yes, it will work on 25xx, but when you have a peer drop it will kill your
> 2501 big time. The CPU on the 25xx is used to forward packets and do the
> BGP4. If you want to try it with 25xx it will work, but I would go with a
> 4xxx router.
I would have to say my experience is different for cpe stub
routers for customers that don't need full tables.
There is no question that at least a 4xxx should be used for any
sizable bgp mesh. However, I used 2500 routers for limited BGP
peers at the periphery of my network with no problems whatsoever. I
just fed my IGP routes to the 2500 via iBGP, peer w/ the transit AS
and take just ^ASNUMBER$ from them. Worked well, no reliability
problems, no cpu problems. Obviously this only works if the
customer doesn't need full tables.
If we were talking about meshing full tables, or multiple peers,
then I'd very much suggest using a 4xxx or 7xxx.
-alan
--
alan@mindvision.com
Not Employed Networking, Inc.
