An experimental study of video conferencing over the Internet

Abstract

Examines the feasibility of conducting a video conference over the Internet. For the study the authors transmitted packets at a fixed rate between two nodes. The question examined is how to determine the correct playout rate and the initial buffering required to ensure a 'good' reception. Two different algorithms to determine the playout rate are proposed and the authors examine their behavior using five different data sets that were collected at different times. The results indicate that the playout rate should be slower (by a fraction) than the packet generation rate and that, surprisingly, the initial prebuffering is irrelevant.