In Search of an API for Scalable Reliable Multicast
Jim Gemmell, Jörg Liebeherr, Dave Bassett · 1997
this paper we explore the issues involved with a reliable multicast protocol for the Internet that can scale to millions of receivers. Ideally, we would like a protocol and API for reliable, scalable multicast that would be universally useful to all such multicast applications. Some would argue that scalable reliable multicast must be implemented at the application level, i.e., that it is not possible to offer multicast reliability as a separate service that is not integrated into the application [FLO95]. To a certain extent this argument has merit. The widely varying requirements of different applications for a scalable reliable multicast service are broad enough to prohibit a generalpurpose solution. However, we will show that it is possible to provide a service that offers mechanisms for reliability which are useful to a sizeable category of applications. In this paper, we define an API for reliable, scalable multicast, while only making minimal assumptions about the protocol. Our goal is to have an API that is useful for experimenting with different scalable protocols. The API that we propose could equally use another protocol to achieve reliable multicast. Defining an "on the wire" protocol for scalable reliable multicast is important, and should be addressed in order to allow interoperability. However, our focus here is with the semantics of a scalable reliable multicast service, regardless of the protocol for implementing it. As a point of departure, we will consider especially the reliable multicast used in the MBONE wb whiteboard tool, called SRM [FLO95].1 SRM has been proven to work well for as many as 1000 participants spread all over the world. However, we are not concerned so much with the specifics of this scheme, as with its general properties, i.e., that...