Error control for continuous media and large-scale multicast applications

One of the key elements that has contributed to the runaway success of the Internet today is its ability to support a wide variety of applications with often conflicting requirements. The Internet, however, offers only a best-effort service model which means that packets can be lost. Applications requiring reliable transmission usually employ a transport protocol to provide reliability. The most common reliable transport protocol today is TCP which provides total reliability at the expense of latency, a trade-off that works well for applications like file transfer. File transfer, however, is just one class of applications. Other classes such as continuous media, transaction-oriented and multicast applications have different, sometimes opposing requirements than file transfer, and thus need different error control mechanisms. In this thesis, we present error control mechanisms for two important classes, namely (a) interactive continuous media (CM), and (b) large-scale multicast applications. The former include applications like teleconferencing, video on demand and visualization; the latter include Internet radio and movies, web and software updates and distributed interactive simulation. In this thesis we have explored retransmission as a possibility for providing error control for interactive CM applications. We have designed, implemented, and evaluated a retransmission-based error control scheme for CM applications, which aims to provide the best possible reliability at the lowest cost, without violating the application's timing constraints. We have enhanced selective-repeat retransmission with: (1) playout buffering to increase the time available for recovery, (2) gap-based rather than timer-based loss detection to minimize loss detection latency, (3) implicit expiration of sender retransmission buffers to eliminate acknowledgments, (4) conditional retransmission requests to avoid triggering late, unnecessary retransmissions, and (5) delivery of data integrity information to the application to aid in concealment. We made the observation that forwarding and error control are two clearly separable components, and great benefits can be realized by decoupling and placing each one where it is more beneficial: the forwarding component at the routers, and the error control component at the receivers. The separation is very clean, and does not violate any layering principles. The result of this separation is a set of simple router forwarding services called Light-weight Multicast Services (LMS). These services do not require examining or storing packets, but greatly simplify multicast control protocols. (Abstract shortened by UMI.)