[c-nsp] Full net table too large for Sup720 already?

[Gary Stanley]

At 11:58 PM 10/26/2007, jim bartus wrote:
>I don't claim to be an expert but I looked into this before and here's
>what I found:
>http://www.nanog.org/mtg-0702/presentations/fib-desilva.pdf
>
>check out page 8, page 10, and the first bullet point on page 15.
>
>Page 10 says the limit on a 3B is 192k by default but can be tweaked 
>up to 239K.
>
>According to the "Weekly Routing Table Report" cisco posts to nanog
>the number currently out there in a full internet bgp feed is 234760.
>http://thyme.apnic.net/ap-data/2007/10/20/0400/weekly
>
>So if I understand that correctly a default 720-3B will already
>overflow and one that is configured to support the max 239K has less
>than 5K left of headroom.
>
>poking through those bgp reports here's the trend:
>2007-05-01:     219238
>2007-06-01:     221952  +2624
>2007-07-01:     224395  +2443
>2007-08-01:     227097  +2702
>2007-09-01:     229742  +2645
>2007-10-01:     233290  +3548
>
>I read that as saying 3B owners have less than 2 months left until
>they can't fit a full bgp table in tcam.  Thats not counting any
>local/igp routes or arp entries.


If you really need to take full tables, just filter out /24's and 
make sure you have a default route being sent from your upstream for 
a 'catchall' on the aforementioned filter list. That will cut tcam 
usage in half, since half of the entries on the routing tables are 
/24's. I do this on some of our gear and it works pretty well.