An empirical study of real audio traffic

The delivery of multimedia content is a facet of Internet traffic that is rapidly growing in importance. The new generation of World Wide Web sites rely heavily on extensive multimedia content such as graphics, sound, music and video to attract and retain visitors. While there have been extensive studies on the growth and effects of hyper text transfer protocol (HTTP) traffic used on the Web, little or no work has been performed in analyzing streaming multimedia traffic. We present the results of a brief study to examine the traffic emanating from a popular Internet audio service using the RealAudio program. We found protocol distributions that show a bias towards non-TCP friendly protocols. In addition, we observed consistencies in audio traffic packet sizes and data rate patterns may be useful as a tool for identifying audio data flows. Our results show that audio flows exhibit significant consistency in data rates and are considerably more persistent than HTTP connections.