[c-nsp] Growing BGP tables

Neil J. McRae neil at colt.net
Mon Nov 22 04:52:34 EST 2004

Engine 0 and Engine 1 have this problem [and enough other problems
that you should pull them out asap they are totally useless cards
same is true for Engine 2 also [esp the gige card] IMO]. 

Early GRP cards have a 256M limit, mostly A's but some B's also we've
found. The good news is that the PRP-2 seems to work as a direct
replacement with 12.0-27.S3 without any issues [at least in our env].

Remember to note that the cards suffer from memory fragmentation quite
badly and you might think you have a window to get them swapped, you
should be doing this now. When you get clowns at random ISP
trying to be #1 in the CIDR-Report and de-aggregating all routes
it can cause real issues. Here is what we have found in our testing
of this issue:

1: Router is running fine and has a decent chunk of free mem.
2: StupidISP.NET de-aggs
3: Router goes "oh I don't have a big enough free() to malloc()
so I better switch of CEF" [and when that happens you can only login
to the box from the GRP ethernet port or the console!!!]
4: Router changes primary function from router to electrical
discharging device! Super!
4a: Other routers in network decide to "down tools" in sympathy.

Also the Flexwan 1 card on the 7609 has the same issue and 
is equally good at electrical discharge.


> -----Original Message-----
> From: cisco-nsp-bounces at puck.nether.net 
> [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Ian Dickinson
> Sent: 22 November 2004 09:13
> To: Ryan O'Connell
> Cc: Vincent De Keyzer; Gert Doering; 'cisco-nsp'
> Subject: Re: [c-nsp] Growing BGP tables
> Ryan O'Connell wrote:
> > I've also heard (but been unable to verify) that Engine 0 
> and 1 line 
> > cards for the GSR won't be able to take a full routing table past a 
> > certain point, which I know if it's true is going to cause a lot of 
> > larger ISPs issues.
> Will possibly happen around the same time as a VIP4-80 is out of time.
> All the 0 and 1 beasties I've seen personally have 256Mb Route Memory.
> It's a short while yet, and will happen eventually, but you 
> can probably kill the prefix-load easily enough by changing 
> usage patterns, as long as you have new iron for the 
> full-table work.  I have some of these older cards running 
> with 200Mb free earning good revenue...
> But the NPE-300 and 3660 et al will hit this problem first and soon.
> --
> Ian Dickinson
> Development Engineer
> ian.dickinson at pipex.net
> http://www.pipex.net
> This e-mail is subject to: http://www.pipex.net/disclaimer.html
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net 
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/

More information about the cisco-nsp mailing list