IP multicast - best practice

[ivarstrandberg]

Hi. I'm setting up IP Multicast for the first time (I'm not experienced with Multicast at all), and I'd like a few pointers in how to set it up in the best way. Attached is a sketch of the network as it is now.

In my lab, I've managed to get it to work: the IOS switches seems to have multicast turned on by default. To get the routers to work, I just entered "ip multicast-routing" globally and "ip pim dense-mode" on the participating interfaces.

What I like to know, is how to set up multicast in my network, so only the clients on 10.1.20.0/24 get the multicast traffic from 10.1.10.37. In other words: as little noise as possible :-)

[Giuseppe Larosa]

Hello Ivar,

test in your lab PIM sparse mode is the mode that provides better control.

>> is how to set up multicast in my network, so only the clients on 10.1.20.0/24 get the multicast traffic from 10.1.10.37. In other words: as little noise as possible :-)

Sparse Mode

Hope to help

Giuseppe

[ivarstrandberg]

When I set all four interfaces to Sparse Mode, nothing happens. Do I need to use RP or something when using Sparse?

[Giuseppe Larosa]

Hello Ivar,

>> Do I need to use RP or something when using Sparse?

yes indeed sorry I didn't write it on first post and also IGMP snooping on layer2 switches is important (may be more important in a campus environment)

you can check RP with:

conf t

ip pim rp

on L3 switch

to verify:

sh ip pim rp mapping

to check IGMP snooping that should be enabled by default

sh ip igmp snooping

see if it says enabled on clients' vlan

edit:

I've seen the network diagram

you need ip pim rp on all L3 devices on path

Hope to help

Giuseppe

[Jon Marshall · Accepted solution]

Just to add to Giuseppe's/Joseph's posts.

Because you only have a small network you can use static RP. So assuming you choose the 6500 as the RP

1) on the 6500 and 7200 add this command in global config mode -

ip pim rp-address 10.74.2.5

2) Under each of the interfaces on the 6500 and 4500 add

"ip pim sparse-mode"

3) As already noted enable IGMP snooping on all your L2 switches.

4) Make sure that the TTL of the multicast packets are at least set to 3

Jon

[ivarstrandberg]

Just the info I was looking for, thanks :-)

One more question though: How can I make sure that the TTL of the multicast packets are set to at least 3? :)

[Jon Marshall]

"How can I make sure that the TTL of the multicast packets are set to at least 3? :)"

That's done within the application itself ie. nothing to do with the network setup as such. Chances are you will be fine but some multicast apps set the TTL to 1 which means it won't go across routed interfaces. There should be some way within the app to tell it to increase the TTL if that is the case.

Jon

[Joseph W. Doherty]

As Jon notes, it should be set by the application, although I recall most hosts default to something like 32, 64 or 128. Whatever; if dense-mode has been working, i.e. TTL has been sufficient, then so it's likely so should sparse-mode. (Reason likely - sparse-mode might introduce [initially/permanently] more hops between source and receiver. [But we're now getting much deeper into multicast then you want, or need at the moment, to know. ;) ])

[Joseph W. Doherty]

Besides Giuseppe's recommendation for using sparse-mode rather than dense-mode to reduce the needless distribution of multicast-traffic, on your switches, if they support IGMP snooping, you would want to insure it's enabled too, to stop the needless distribution of multicast to ports on a network segment.

BTW, sparse-mode is little more complex to configure than dense-mode since a RP needs to be defined.

Also, if your network is such where multiple routers share a network segment, PIM snooping might be used to supress multicast traffic to router ports that don't need the traffic.

PS:

"Attached is a sketch of the network as it is now."

Appears to be missing.

[ivarstrandberg]

Hmm...weird, I did attach it. But anyway: http://www.argaste.com/files/multicast.png

[cbeswick]

Hi,

There is maybe a another way to do this and that is to use auto-rp. By using this feature you do not need to manually configure a rendevous point for every multicast group, instead the network takes care of it automatically.

To enable this you configure a minimum of 2 routers to announce themselves as candidate rendevous points, and a minimum of 2 routers to act as mapping agents. The mapping agents then choose between the 2 candidate routers and automatically assign a router to a multicast group.

It sounds complicated, but once setup you can just leave it alone and let it do its thing. We have been using this configuration for over 3 years now without issue. We also use IGMP snooping on all access layer switches and PIM snooping on core / distribution devices.

In order to use this you need to configure ip pim sparse-dense-mode on your backbone. This is needed for auto-rp to function.

The BCMSN book used to discuss this configuration, but I do not know if it is still included on the course.

HTH.

[Giuseppe Larosa]

Hello Chris,

good note

there are only three routers in the picture so using manual RP setting is acceptable here.

Hope to help

Giuseppe