[c-nsp] Multicast state behavior differences between 7600 and ASR9k

[Phil Mayers]

On 01/03/11 15:41, John Neiberger wrote:

> sources with no receivers. The router must already be keeping track of
> those sources, though. If an IGMP or PIM join arrives, the router will
> know to start forwarding that traffic. So, it seems as if the router
> has been keeping an eye on the multicast traffic on that port. Why not
> have a simple command that shows it?

I'm not sure your conclusion that the router "must already be keeping 
track" follows for SSM though.

Why can't the hardware be programmed something like the following:

s=1.2.3.4/32 g=233.x.y.x   oif=[a,b,c]
s=5.6.7.8/32 g=232.a.b.c   oif=[d,e,f]
s=0.0.0.0/0  g=232.0.0.0/8 drop
s=0.0.0.0/0  g=224.0.0.0/4 punt ratelimit=<assert limit>

i.e. for the SSM group range(s) configured, program in "drop" rather 
than "leak to CPU for PIM register" That is how I'd implement it; for 
SSM when punt-to-CPU is not *needed*, don't punt.

For ASM you have to initially punt to CPU and *have* to keep the state 
to manage the register & register-stop, so there I would expect to see 
state for groups with no members.

Obviously it depends on the nature of the forwarding hardware.