Multimedia transport protocol and multicast communication

The emerging of high speed networks provides the infrastructure for handling a wide set of new applications, covering distributed multimedia cooperative features. In this framework, multipeer communication is a major service/protocol issue because most of these applications exhibit the need for such a functionality. Besides many functions and mechanisms, the semantics of transport layer multicast conversation has to be defined. We show that a new service semantic called statistical-reliable, in which the reliability is enforced when desired by applications, is of major interest for the high-performance networking paradigm. This new reliability semantic can be implemented using a protocol akin to the XTP bucket algorithm. The interest of the developed protocol is that 1) it can perform efficiently on top of wide area (for instance ATM-based) networks as well as LANs, 2) it provides a continuous reliability service from best-effort to all-reliable as a function of the application requirements, 3) it is able to handle group size from small to very large with only a slight increase of overhead, 4) it works for unicast communication. These functionalities provide an interesting and efficient protocol for covering the need of distributed multimedia applications.