Author_Institution :
Sun MicroSyst. Inc., Menlo Park, CA, USA
Abstract :
Given about 15 year´s experience with IP multicast, it is time to reevaluate some of the basic assumptions. The original IP multicast model copied, as closely as possible, the Ethernet multicast model. The Ethernet multicast model was well suited for both the technology (shared bus LAN) and the types of applications that would use layer 2 multicast. However, the IP multicast model did not prove to be successful. This paper argues that the types of uses of layer 3 multicast differ so much from the types of uses of layer 2 multicast, that it was not wise to attempt to make the model the same. Also, the paper explains the difference in technology between Ethernet and IP that made that model impractical to implement for layer 3. Finally, the paper shows that no single model for IP multicast is ideal. It presents two types of applications, and shows how each of those require a different multicast model. This shows that IP multicast requires at least two different models, neither of which is the original IP multicast model. This paper describes those two types of applications, together with implementable models that would be well suited for them.