Multicast

[CiscoIPMAN]

I am trying to get an imaging server (Microsoft WDS) to be able to image clients on a different subnet than the server. A few questions maybe you guys could help me with.

1. From my readings the L2 switches that the clients are connected to must have the global command " IP IGMP", is that correct?

2. All the L3 switches that are in the path to the L2 switch must have multicast routing turned on? They must have IP service image?

3. Is turning on multicast routing enough or do I need to set up the RP and other configuration?

Thanks for any help

[Giuseppe Larosa]

Hello @CiscoIPMAN ,

L2 switches will need to have

ip igmp snooping

enabled

L3 switches need

ip multicast-routing

+

ip pim sparse-dense mode

under each L3 interface ( routed or SVI Vlan inteface)

3) an RP is not strictly required but you can configure one manually using

ip pim  rp <address>

the RP address should be a loopback interface on a L3 switch centrally located.

This can provide some benefits if you have a lot of multicast traffic

Hope to help

Giuseppe

[CiscoIPMAN]

Thank you Giuseppe!

You said each L3 interface or SVI, is this on every router or switch that the multicast traffic will go through? So if I do a tracroute from the server network to where the clients are located those switches need to have ip multicast-routing and ip pim sparse-dense mode?

Is there a potential to flooding the network with multicast traffic?

Thanks

[Giuseppe Larosa]

Hello @CiscoIPMAN ,

Yes ip pim sparse-dense mode has to be enabled on all L3 interfaces on the end to end path ( and also on possible alternate paths if any for redundancy and fault tolerance)

>> Is there a potential to flooding the network with multicast traffic?

Actually no because even dense mode that used a Flood and Prune approach is able to remove vlans that do not need to receive a multicast stream,  but the price to pay is a state of Pruned of 3 minutes . Before the expiration of this 3 minutes timer the downstream router ML switch that does not need to reiceve traffic for a stream will re-send a PIM prune for the group G avoiding re-flooding.

This means that each node will need to keep a state machine for each group Gx to list what interfaces to send it out (forwarding ) and which to stay pruned.

So dense mode requires more resources then Sparse mode where the idea is the opposite an Explicit Join is needed to receive traffic in terms of memory on multicast nodes ( so called state)

Hope to help

Giuseppe