You have the option of configuring a specific group (Class D address) or allowing the server to generate one according to your choice of addressing / scope.
The advantage of assigning one is that you can then "tighten down" your routers, switches, and firewalls for security or traffic flows / traffic engineering.
The drawback is the administration. The degree of pain will vary according to how often you create programs (and it's not really that painful, once you get the hang of it).
You routers and (perhaps) your switches should be configured for proper multicast handling. The routers usually get set up for PIM (sparse or sparse-dense), the switches get set up for CGMP (prunes back ports that have no group members).
The clients will send a request to participate as a member of a particular group (learned from the "Program Guide" or similar mechanism). The routers (and usually the switches) will acknowledge the request, and light up the multicast stream for that branch/segment. The nearest router will occasionally poll the clients to make sure they're still listening....if they don't respond tothe poll, the router stops the multicast stream (when there is no response to the poll).
This is a very basic overview; standard rules apply: "Some information bent, folded, spindled or mutilated for the sake of brevity."
FWIW
Scott