Illustrated TCP/IP by Matthew G. Naugle Wiley Computer Publishing, John Wiley & Sons, Inc. ISBN: 0471196568   Pub Date: 11/01/98

Chapter 303
Multicast Conclusion

Multicasting is coming and it is taking many forms on an internet or the Internet. We have voice, video, and data multicasting. There is one common protocol message format, and that is IGMP. PIM and DVMRP use this framing for their messages as well. MOSPF uses its own LSA for communicating between routers. However, all multicast applications conform to IGMP to register themselves on the local subnet.

All of the protocols have their own advantages and disadvantages. MOSPF requires that you are running OSPF. A RIPv1 or RIPv2 network cannot simply install MOSPF. If the site chooses to use RIP, it must use PIM or DVMRP. MOSPF converges instantly, whereas PIM and DVMRP may have routing loops during a slow convergence. However, installing MOSPF is not a simple task, for MOSPF is OSPF with multicast extensions; therefore, all of the rules with OSPF still apply. MOSPF does not support tunnels. It expects some type of other protocol such as DVMRP to be running on the M-ASBR routers. DVMRP does not scale well and neither does PIM without some tuning. DVMRP and MOSPF work better in densely populated environments, whereas PIM has two modes of operation: sparse and dense mode. PIM allows for duplicate packets in dense mode, in favor of a simpler protocol that is not dependent on a unicast routing mechanism. You really must consider all the options before placing a multicast protocol on your network.

One of the things you must think about with multicast is not only receiving data, but how to respond that data. Yes, multicasting does allow for one station to send one data-stream to be received by literally tens of thousands of receivers. But what if there is a need for acknowledgment of that data? This poses a considerable problem: one sender, multiple receivers. If the number of receivers is 2000 and they all need to send some type of Receiver Status message back to the sender, the sender must be able to handle this type of back-channel information flow—especially if the status is long. The inability to handle this is called implosion, during which the originator is overrun by the back channel.