[c-nsp] L3VPN VPNv4 NLRI - Route Reflector Scaling

[Mark Tinka]

On Monday 24 March 2008, Oliver Boehmer (oboehmer) wrote:

> Well, most of the L3VPN deployments I'm aware of (which
> includes some very large SPs) still use a single iBGP
> mesh of dedicated VPNv4 RRs, some flat, some using
> hierarchical RR structure. RR partioning via rr-group or
> using other means is rarely done as the scalability
> requirements are still able to be handled by the simpler
> design. I guess once you reach 500.000 vpnv4 prefixes or
> more, RR partitioning comes into play, with the caveats
> you've mentioned. What are your requirements?

We would like to build scalability into the network for 
VPNv4 route reflected NLRI early on so that there is little 
to change when we start seeing that number of prefixes.

At this time, we see simple route reflectors handling all 
address families as the way to start. As the network scale, 
doing the same on dedicated VPNv4 route reflectors seems 
logical.

Beyond that is what we are thinking about. We might be able 
to live with additional routing information at the PE 
routers initially, but it would be an area of concern at 
scale.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 832 bytes
Desc: This is a digitally signed message part.
Url : https://puck.nether.net/pipermail/cisco-nsp/attachments/20080324/552e4f3f/attachment.bin