Characteristics of Streaming Media Stored on the Web

The increasing power and connectivity of today’s computers have spurred the growth in streaming audio and video available on the Internet through the Web. While there is substantial research characterizing the performance of streaming media and characterizing documents stored on theWeb, there have been few studies characterizing streaming audio and video stored on the Web. We crawled over 17 millionWeb pages from key geographic locations and extracted nearly 30,000 streaming audio and video clips for analysis. Using custom built tools, we analyzed the characteristics of these multimedia objects, determining such information as media type, encoding format, playout duration, bitrate, resolution, and codec. We find proprietary audio and video formats dominate all multimedia content, primarily content by RealNetworks followed next by Microsoft and with Apple following just behind MP3. The playout durations of streaming audio and video clips are long-tailed, suggesting streaming media may contribute self-similar traffic on the Internet. More than half of all streaming media clips on the Web are video, with 90% of videos targeted for broadband connections. Video resolutions are considerably smaller than typical monitor resolutions, implying that video bitrates, which are directly related to resolutions, have enormous potential to increase. The detailed results from this study should be useful for future studies characterizing the performance of streaming media on the Web and also valuable for those interested in generating more accurate Internet traffic simulations