Hello Rmundy,
you can start from this multicast tutorial
http://www.cisco.com/en/US/docs/ios/solutions_docs/ip_multicast/White_papers/mcst_ovr.html
multicast addresses represent sessions for example a videoconference and they are associated to a content that may be of interest for multiple receivers. Think of it like a TV channel or a radio channel. The stream has to be carried in all network segments where at least one receiver is present and has signaled his interest.
There are no concepts of multicast subnets or locations. Each multicast address represents a different channel.
Interaction between routers and receivers is managed by IGMP protocol (see the tutorial)
Another protocol PIM that has different flavors ( sparse mode, dense mode, bidirectional or source specific sparse mode) is used in the network infrastructure.
Multicast routing has to be enabled on all network devices from multicast stream source to receivers.
Multicast routing builds dynamically a distribution tree from source ( rooted at source) to interested destinations.
The objective of multicast routing is to minimize the number of copies of the stream over the network to minimize bandwidth usage.
So replication is performed when needed to send the stream in different directions (for example to two different remote sites).
The basic check that each multicast router device performs is called Reverse path check: a multicast packet is considered valid if it is received on the interface pointing to the source IP address, this means that PIM actually works on IP unicast routing of multicast sources.
This is the objective of PIM.
In your case having many locations and with different types of devices and link bandwidths you should deploy IP PIM sparse mode.
PIM sparse mode is advantageous because it does not waste bandwidth: the multicast stream is propagated only to remote sites that have sent and periodically send an explicit PIM join for the multicast group.
PIM SM requires the use of a central arbiter called rendes-vous point and builds the distribution tree using a shared tree rooted at RP router.
(In a second time a distribution tree rooted at the source is built as a form of optimization)
Knowledge of the RP address has to be given to all multicast nodes. In small networks this can be manually configured on all non RP devices. In a network like yours you can take advantage of bootstrap protocol or auto-RP.
The basic configuration steps for manually configured RP are:
in global configuration:
ip multicast-routing
ip pim rp-address x.x.x.x
x.x.x.x is a unicast address of the RP device it is usually a loopback interface advertised in the IGP you use ( OSPF. EIGRP or other)
on all interfaces pointing to network devices or to which receivers are connected (because also IGMP is automatically started when enabling PIM on the interface this is the key point)
interface type x/y
ip pim sparse-dense-mode
After you test this and you understand it you can move to the use of bootstrap protocol (recommended) or auto-RP
Hope to help
Giuseppe
