[c-nsp] full routing table

[Alex Howells]

>>
>> Thanks guys :)  Was just pondering whether a Catalyst 4948 would be good
>> enough for deployment with two partial feeds, as 76xx series is somewhat
>> expensive for that particular project!
>>
>> Guessing the FIB on it will be the limiting factor.
> 
> Very very partial feeds.  The 4948 is limited 32,000 routes.  Even the 
> 4900M is limited to 200,000.
> 
> http://tinyurl.com/2zd3v4
> 
> If you want to carry full tables you'd be better off looking at routers 
> and not switches.  I have 3800s and 2800s with full tables (even had 
> some 3600s).  If this is a new purchase then I'd recommend a 7200 such 
> as the 7201.

Asking purely out of 'curiosity' because a friend who works for an 
up-and-coming social network aggregator is needing to expand his 
network, starting to take on their own feeds, etc.

They've acquired a couple of Catalyst 4948's on the advice from a 
network engineer that they'll take two full tables.

Whilst this is indeed true, they wouldn't be able to *use* the full 
tables so I was just double-checking my information - thanks!

Speaking from work-perspective, we've been using Linux + Quagga up until 
'07 when we acquired a 7600 series - they're bloody marvellous.. 
Deploying another 3 x 7600 series to completely obliterate Linux/BSD 
from our core routing layer in the next 1-2 months, hopping with 
excitement actually as they should arrive in 7-10 days :P

show memory statistics        [pruned a bit]
                 Total(b)     Used(b)     Free(b)
Processor      928148912   358931896   569217016

FIB TCAM usage:                     Total        Used       %Used
      72 bits (IPv4, MPLS, EoM)     524288      246285         47%
     144 bits (IP mcast, IPv6)      262144        1738          1%


Seems that two feeds with ~240k routes and ~200k routes respectively 
plus some partial/peering use about 215MB (bgp process).