Multicasting Question over L3

Unanswered Question
Sep 19th, 2008

We are working on implementing a multicast solution on our LAN. We would like to push multicast out to our closet switches which are connected via L3 connections. What in your opinions is the best way to go about this. When we were L2 we used IP pim sparse-dense-mode with cgmp but it was just on one vlan which pretty much encompassed all of our workstation IP's. Now as stated, we are L3 to all of our closet switches and have seperate vlans and IP schemes for each switch. Basically, how do you push multicast over L3? Do you have to do anything to the routed links?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (5 ratings)
Jon Marshall Fri, 09/19/2008 - 06:53

At it's most basic level.

1) On each L3 device enable "ip multicast-routing"

2) On each interface you want to route multicast traffic from/to

int fa0/1

ip pim sparse-mode-dense

3) Define your RP's etc. You presumably know how to do this as you already had pim-sparse-dense mode running ?

That will do it. IGMP will be enabled for those vlans you have done step 2 for.


dhopper82 Fri, 09/19/2008 - 11:34

So, going off that link and without doing the loopback address (for the time being) my configuration would look like this...

On core switch:

ip multicast-routing

ip pim rp-address (the gateway of the vlan (60) that the multicasting device is on)

interface vlan 60

ip pim sparse-dense-mode

interface tengigabit 12/4 (layer 3 link)

ip address

ip pim sparse-dense-mode

On Closet Switch:

Inteface tengigabit 13/1 (layer 3 link)

ip address

ip pim sparse-dense-mode

ip multicast-routing

ip pim rp-address

interface vlan 800

ip pim sparse-dense-mode

I used the address on the vlan as the RP because I have multiple L3 links that I will need to send multicast to.

dhopper82 Fri, 09/19/2008 - 07:14

Thanks guys! That's was I was thinking I needed to do on the L3 interface. I'm going to try it and see if it works.

My next question is this. The RP would be a vlan on our core router (I think since it's closest to the source). We are able to get to this vlan obviously from our closet switches. On my L3 interface would you reccomend that I put the ip address of the RP as the interface address of that vlan or, would I use the L3 address on the other end (the one attached to the core)?

Jon Marshall Fri, 09/19/2008 - 07:18

In a L3 environment i would do as Michael suggests and create a loopback interface and use this as the RP address.


cowetacoit Fri, 09/19/2008 - 09:08

Just create a loopback interface on the core.

interface loopback0

ip address x.x.x.x

Then on your routed-access layer switches, configure the RP with the loopback IP.


ip multicast-routing

ip pim rp-address x.x.x.x

Refer to the Cisco documentation i posted previously. That's how i learned to configure multicast. I use to think it was very complicated but in fact it is pretty easy to grasp.

dhopper82 Fri, 09/19/2008 - 11:37

The only reason I'm not totally grasping the loopback is because the cores themselves are vlan'd off. We're L3 to our closet switches which again have vlans specific to the switch.

Jon Marshall Fri, 09/19/2008 - 13:11

If your access-layer is connected to your core/distro via L3 links then you must have some routing going on between them, whether this be static routes on the access-layer switches pointing to the core or a dynamic routing protocol.

So all you have to do is advertise your loopback address to the access-layer switches via either

1) Your dynamic routing protocol

2) a static route on access-layer switches

If you are using a default-route on the access-layer switches pointing to the core/distro then you don't need to add anything routing wise.


paul.matthews Sat, 09/20/2008 - 02:23

If you are running L3 across the board, and there will be a router between any device and your core routers, I would use anycast-RP.

Basically have the same IP address on the loopback of both core routers, and make sure one is preferred through the L3 infrastructure. If you cannot make one preferred use MSDP to get them to exchange information about sources.

All other L3 devices have that (shared) loopback address configured as the RP address, and PIM enabled on routing interfaces.

It does not matter about the RP being near the source, as it is by default just a meeting point. As soon as traffic gets to the router that has the receiver, the receiver will switch to the source tree rather than continue to use the shared tree.

dhopper82 Mon, 09/22/2008 - 07:28

Thanks! That was the explanation I needed for the loopback address. I'm still debating on whether or not I need it as the Multicast Source is the only thing on the vlan at the moment. Therefore, I'm not too worried about additional traffic on it.

paul.matthews Mon, 09/22/2008 - 07:58

I'd do it properly. Multicast will probably only grow, and then you have the job of changing a live network. It is not difficult to do properly.

dhopper82 Mon, 09/22/2008 - 09:36

Good point. If we ever get another multicast device and want to put it on another vlan, which is highly likely, having one encompassing address would be nice.


This Discussion