All things related to multicast

[Peter John Hill <>]

We have two separate layer 2 cores. All of our distribution/building 
routers have an interface to each core (two per router, one per core). 
Multicast does indeed go everywhere on the cores. We have gig 
everywhere, so it is not much of a problem. We can't turn on igmp 
snooping on the cores, as that would totally screw up pim and igmp. (we 
know from experience). So currently we deal with it. We will eventually 
add another set of cores for layer 3 connections between our routers. 
We will run multicast over those links. The layer 2 core will only 
continue to exist to allow us to have a few subnets exist across the 
entire network (wireless for example).

So basically, we don't worry about it. I can see that it could be an 
issue if you have some tighter links to the cores. If there are only a 
few smaller links, perhaps you could start routing them first. We are 
using Cisco 6509's sup1a/msfc 1s in our core. We are routing to our 
egress from the cores, and all is good. If you don't have the 
capability to route on your core boxes now, perhaps you might consider 
getting that ability sooner rather than later.

We are lucky that we have dark fiber and short distances between all of 
our buildings (all within limits of gigE over single mode) I am not 
sure how big your campus is, and if you have other constraints. Hope 
this helps.

Good luck!

Peter Hill
Network Engineer
Carnegie Mellon University

On Wednesday, September 25, 2002, at 02:40  PM, Mark Boolootian wrote:

> I'm interested in knowing how folks deal with multicast across a
> layer 2 core.  Our campus backbone currently has a layer two core which
> is treated as a common backbone subnet.  While that makes traceroutes 
> look
> nice, multicast traffic floods and all the backbone router interfaces 
> see
> it, whether they need to or not.
>
> As we roll out gigE in the backbone, the topology is moving towards a
> collection of point-to-point links, so the multicast flooding over a
> layer 2 core problem will go away.  Nonetheless, I'm interested in 
> knowing
> how others have dealt with this issue.  It would seem your choices are
>
>   o Overlay the layer 2 core with layer 3 point-to-point links
>
>   o Run the appropriate protocol for mitigating this problem.  Cisco 
> has
>     RGMP, which sounds like CGMP all over again.  Foundry, and perhaps
>     others, is doing PIM snooping, which sounds like the right way to 
> do
>     it to me.
>
>   o Live with it.
>
> Any experience/comments would be appreciated.
>
> thanks,
> mb
> ---
> Mark Boolootian
> UC Santa Cruz