Re: [pim] MRIB and multiple RP's

[Stig Venaas <stig.venaas at uninett.no>]

John Zwiebel wrote:
> On Nov 11, 2006, at 6:48 PM, Bandar Al-Turaif wrote:
>
> > I know that the RP at it current state does not know any details about 
> > DR’s. My question is, is it possible to modify PIM-SM and make it learn.
> >
> > Our research is about relocating the RP depending on the shared tree 
> > cost, so it is very important for us to know the total cost of tree 
> > (link cost metric, or delay metric).
> >
> >  
> >
> > With my best regards.
> >
> >  
> >
> > Bandar Al-Turaif
> 2 cents:
>
> Of course it is possible.  However you should consider that a bidir RP 
> does not require a physical
> system to be used, its just any arbitrary point in the network.  This is 
> both good and bad for what you
> want.  Good, because it makes RP relocation simple: there's no need to 
> physically reconfigure a router.
> Bad because there's no central place to collect the information you are 
> interested in.
>
> Of course, you are really interested in just Sparse-mode and not bidir.

IMHO this exercise is much more interesting for bidir. For SM you would
usually switch to SPTs anyway so that RP location does not matter much.

> IMHO what you want is
> a "two-stage" effort.  The source and/or receiver DR registers with a 
> "RP-allocation server" (ie something that
> does the calculation to determine the best location of the RP) and then 
> a redistribution of RP
> information for the specific multicast group.  (ie a /32)

Yes, sounds reasonable.
> FWIW, the bootstrap router protocol could quickly redistribute the /32 
> RP map within an enterprise.

Would be kind of interesting to use BSR and assign the different C-RPs
weights according to the shared tree cost for each of them for each
group. I'm not sure you want to move the RP when someone new joins the
group though. I mean, the optimum RP location might change when someone
new joins, but moving the RP and hence the shared tree, while data is
being forwarded is likely to cause problems. This might be ok if
forwarding is done on SPTs, but then why move the RP...

BTW, the RP should be able to determine the cost for the first hop
routers (source DRs), learning their unicast addresses from the
PIM registers. If all receivers also are sources...

Stig

> However, doing this across PIM domains would be problematic.
>
> So, you have a potentially good idea, but it depends on the continued 
> deployment of sparse-mode
> (ie some, like me, feel sparse-mode is no longer useful and should be 
> replaced by SSM procedures)
> and you need to define a protocol that is going to be rather complex, 
> especially if it is used interdomain.
>
> At this point, I do not feel you'll get any vendor to implement it.  
> Just an opinion.
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> pim mailing list
> pim at ietf.org
> https://www1.ietf.org/mailman/listinfo/pim



_______________________________________________
pim mailing list
pim at ietf.org
https://www1.ietf.org/mailman/listinfo/pim