Non Blocking Multicast
My current company has gone all in with Cisco on their IPFM (IP Fabric for Media) using a “Spine/Leaf” solution. I put Spine/Leaf in quotes because it definitely isn’t traditional. We have a chassis acting as the Spine with all of the Leafs attached. . . . which is pretty much just a standard deployment of a core switch with other switches attached to it.
Anyway the real reason we are doing this is for the NBM (Non Blocking Multicast) function. To understand what this is you need to understand how multicast works today. Today if you have a link with 10 gig link with 100 multicast groups on it and that link is at saturation but working fine, when you add the 101’s multicast it will break all the other 100 that were working perfectly fine before until you free up bandwidth on that link.
With non blocking multicast the name is actually a little counter intuitive in this specific example. It would block the 101’s stream because it would know that there isn’t enough bandwidth to continue without affecting other streams. It knows the bandwidth amount due to policies you create. You associate a multicast group with a policy and a bandwidth allocation to the policy. If the calculation is more then what is available it will fail.
In a different example which would be more typical of a Spine/Leaf setup, you will have multiple connections going down to multiple leafs. NBM will distribute the traffic across the links accordingly so you will hopefully never have to block traffic on the link unless you get into over-subscription territory again for some reason.