ON RETRANSMISSION-BASED ERROR CONTROL FOR CONTINUOUS MEDIA TRAFFIC INPACKET-SWITCHING NETWORKS

Distribution of continuous media traffic such as digital audio and vid eo over packet-switching networks has become increasingly feasible due to a number of technology trends leading to powerful desktop computer s and high-speed integrated services networks. Protocols supporting th e transmission of continuous media are already available, In these pro tocols, transmission errors due to packet loss are generally not recov ered. Instead existing protocol designs focus on preventive error cont rol techniques that reduce the impact of losses by adding redundancy, e.g., forward error correction, or by preventing loss of important dat a, e.g., channel coding. The goal of this study is to show that retran smission of continuous media data often is, contrary to conventional w isdom, a viable option in most packet-switching networks. If timely re transmission can be performed with a high probability of success, a re transmission-based approach to error control is attractive because it imposes little overhead on network resources and can be used in conjun ction with preventive error control schemes. Since interactive voice h as the most stringent delay and error requirements, the study focuses on retransmission in packet voice protocols. An end-to-end model of pa cket voice transmission is presented and used to investigate the feasi bility of retransmission for a wide range of network scenarios, The an alytical findings are compared with extrapolations from delay measurem ents of packet voice transmission over a campus backbone network.