[j-nsp] RIB and FIB - Memory for MX with LR

Mark Tinka mark.tinka at seacom.mu
Sun Jun 30 15:29:17 EDT 2013


On Friday, June 28, 2013 06:52:54 AM joel jaeggli wrote:

> which means the bgp rib can grow to fairly outlandish
> size and not crap up the fib associated with that
> logical system.

I wonder how much more room you could gain if you didn't 
even have to install BGP routes in the RIB.

In IOS XE on the ASR1000, Cisco have a feature, so-called, 
BGP Selective Route Download. It prevents routes from 
entering the RIB on dedicated route reflectors if the route 
reflector clients are all setting the NEXT_HOP attribute to 
"self".

This, apparently, provides a theoretical upper limit of 
between 22,000,000 - 24,000,000 million routes that a single 
router with 16GB of RAM can handle in the control plane a 
dedicated route reflector role. 

The number is brought down by one-third, however, if routes 
are allowed to enter the RIB.

I'll be testing this out myself soon.

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/juniper-nsp/attachments/20130630/213f5310/attachment.sig>


More information about the juniper-nsp mailing list