[c-nsp] Growing BGP tables

[Gert Doering]

Hi,

On Fri, Nov 19, 2004 at 12:44:23PM +0100, Vincent De Keyzer wrote:
> I sometimes read scary allusions to a coming catalysm on the Cisco boxes
> while the BGP table increases in size.
> 
> What exactly is the problem?
> 
> More precisely, we are running 7206 VXR with NPE-400 and 256 MB (single eBGP
> feed per machine), how far can we go? What do we have to upgrade before we
> reach this limit?

Our 256 Mb-boxes run at about 50-55 Mb free right now (12.2(18)S, 12.2(25)S
will leave less free mem).  From my gut feeling, I'd say that as soon as
the routing table reaches 200.000 routes, these boxes will run out of RAM,
and subsequently will start flapping BGP sessions, disabling CEF, and 
doing "unplanned" things.

The NPE-400 can go to 512 Mb DRAM, so it's "just" a matter of money.

NPE-225 and NPE-300 can only go to 256 Mb (at least that's what
documented), so the unlucky owners of these will need to get NPEs, and
in some cases, new VXR chassis (NPE-225 sitting in non-VXRs).


Our consequence is that we'll start to filter out large parts of the /24s
cluttering the routing table.  We're certainly not going to invest LOTS
of money just because other people think they can be lazy and not 
aggregate (or because they have some confused ideas about "this is the
only way we can prevent route hijacks").  

We just can't *afford* replacing the existing 7206/NPE-225s that don't
carry that much traffic (peak 20-30 Mbit/s, the NPE-225 does its job
very well here) with new gear - unless "new gear" is driven by 
"customers needing more bandwidth, and paying for it".

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de