All things related to multicast

[Bill Owens <>]

At 17:46 -0500 10/9/00, Bill Nickless wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
>
> At 06:26 PM 10/9/2000 -0400, Bill Owens wrote:
> > The big question, though, is whether it would work at all. Has anyone
> > tried this sort of thing? Does it seem to make sense?
> >
> > Thanks for any help you can provide,
> > Bill.
>
> Yes, the plan you laid out can be made to work.
>
> However, if you're going to go to the trouble of running M-BGP and PIM-SM
> everywhere, I am confused as to what you gain by NOT running MSDP.  You're
> already accepting the burden of PIM-SM protocol operation, which means a
> relatively late version of router code.  You're already accepting the
> burden of M-BGP, which means an additional requirement for memory
> (proportional to the number of NLRI=Multicast routes) in each of your
> routers.  And with both of those you're already accepting a requirement for
> configuration tasks in your backbone core and at each of your customer sites.

Actually, we're not going to run MBGP either; the idea was to have 
just PIM-SM within NYSERNet's own network and our customers'. Some of 
our customers have routers that can't be upgraded to support the full 
troika, others have too much work to do to get to that point quickly, 
etc. This configuration is similar to what a lot of people are 
probably running internally, but they're using PIM and an IGP, where 
we want to use PIM and unicast-only BGP, plus IGP's within the 
customers' networks.

> The most immediate downside that I can see is that you're stuck with a
> single RP for all of your customers.  Problems related to
> single-point-of-failure come immediately to mind.  Without thinking too
> hard about it, I'm afraid you'll have to manually(?) maintain some ugly
> access control lists so your RP knows what prefixes it should advertise.

If the access lists have to contain our customers' prefixes, that 
isn't too bad; we have only a few, and most have just a couple of big 
blocks (they're network old-timers ;)

> The configuration you outline makes this problem much harder
> to solve, since the problem of MSDP SA generation could be inside your
> backbone or at your customer site.

That could be a significant troubleshooting problem. Right now we 
don't have any network monitoring equipment in the backbone to see 
whether sites are doing the right thing. My understanding of PIM is 
too fuzzy to know whether it would be possible to diagnose this by 
looking at the PIM state of the backbone routers that the customers 
connect to. . .

Bill.