Re: [v6ops] draft-ietf-dhc-pd-exclude

[jouni korhonen <[e-mail]>]

Hemant,

On Dec 7, 2011, at 3:40 PM, Hemant Singh (shemant) wrote:

> -----Original Message-----
> From: jouni korhonen [mailto:[e-mail]] 
> Sent: Wednesday, December 07, 2011 1:59 AM
> Cc: Ray Hunter; Ted Lemon; Thomas Narten; [e-mail]
> Subject: Re: [v6ops] draft-ietf-dhc-pd-exclude
>  
>  
> > setup on the DR that is sending an RA with one exclude prefix if the DR has 20K different IPV6 CE routers  as RR’s?  The RA >is multicast and, say, reaches 40K CE routers.   How is the exclude prefix working with the multicast RA?

This would work assuming you then include the pd-exclude only to that single IA_PD that is aggregate of the prefix on the link. The rest of the 19999 RRs' IA_PDs do not need pd-exclude for the prefix.

I setup my home network to mimic a deployment like this just to verify it worked.. i.e. my 'DR' had a Pref::/56 route to 'CE' A. The link where the rest of the routers B etc were, all configure their 'WAN' using SLAAC from Pref::/64 (the excluded prefix).

I think it could be useful to explain this case in pd-exclude draft how and why it actually works according to the conceptual sending algorithm in RFC4861.

- JOuni

>  
> >Is this a real deployment scenario you are designing or already having where you intend to use pd-exclude? My question >regarding this specific example is why you would use RAs & SLAAC to configure 40K CE routers' WAN links attached to a single >link (with one prefix) and then try to apply pd-exclude in the first place?
>  
> I do have a cable deployment where I can have the access concentrator serving 100K PD clients.  The network would like to use the pd-exclude.  The reason I asked my question because one person already said, the DR can use the RA so that the DR can address it’s interface with SLAAC.  There are two choices for such a deployment and that is why I asked the question.  Either the same exclude /64 is given to each of the 100K clients or the network uses unicast RA.  Some IETF document has already defined a unicast.
>  
> Hemant
>