Web Browsing

Characteristics of Streaming Media Stored on the Web

Improvements in the connectivity levels of today's computers have enabled Web users who cross cultural and national boundaries to stream multimedia applications from far away Web servers to browsers on their desktops. Whether it is news, sports, or entertainment clips, the newest generation of Web users expect the convenience of being able to initiate audio and video streams by simply clicking on a browser link. In 2001, Real Networks estimated that 350,000 hours of online entertainmentwas being broadcast each week over the Internet, and this statistic does not include the volume of additional hours downloaded on-demand by Web users around the world.

AIDA (Cooperative Association for Internet Data Analysis) emphasized in 2002 the significant fraction of Internet link capacities that were being allocated to support streaming media applications. Announcements such as RealNetworks' 2003 press release to support the advancement of streaming multimedia applications over wireless cellular networks add to the concern among Internet experts about the ability to support access to streaming media clips that are readily available on the Web. This anxiety over future streaming media applications significantly restricting performance for other Web users has translated into a variety of research papers that propose new network protocols [Floyd et al. 2000; Rejaie et al. 1999] or more sophisticated network router algorithms that seek to lessen the anticipated effect of streaming media [Mahajan et al. 2001; Feng et al. 2001; Cao et al. 2000; Stoica et al. 1998] on Internet performance. Several recent research efforts [Chesire et al. 2001; Wang et al. 2001; Li et al. 2002; Mena and Heidemann 2000; Veloso et al. 2002; Chung et al. 2003; Kuang and Williamson 2002a] have focused on capturing the characteristics of current streaming application behavior to better understand its impact. Only by knowing the relative frequency of commercial streaming products and how they typically stream multimedia traffic can researchers begin to prepare for the next generation of Web users.

Unfortunately, there is little recent published work on specific characteristics of streaming media clips stored on the Web. While there have been studies characterizing Web content measured at the client side [Bray 1996; Woodruff

et al. 1996], there have been no recent studies of the general attributes of streaming media clips stored atWeb servers. In 1997, Acharya and Smith [1998] studied video content stored on the Web by analyzing every video available in the (then popular) Alta Vista search engine. However, the nature of streaming media has changed considerably since that time. For example, Acharya and Smith [1998] found that the Internet could not support real-time streaming given the encoded bitrates and last-mile connection capacities available in 1997. Today, RealNetworks ‘RealPlayer and Microsofts' Media Player, two popular streaming media products [Jupiter Media Metrix 2001] that did not even exist in 1997, RealNetworks' have significantly improved aWeb user's ability to stream multimedia to home computers.

The papers by Ousterhout et al. [1985] and by Baker et al. [1991] proved to be influential in the design of new file systems and distributed file systems because they provided fundamental research on the nature of data stored in file systems and how these files were likely to be accessed. Accessibility to media clips on the Web through a variety of commercial streaming media products has reached such a state that similar studies on the characteristics of streaming media stored on the Web are needed to appreciate the future impact of millions of Web users around the world concurrently streaming freely available stored multimedia clips from remote Web servers to media clients in their homes.

This investigation built customized tools to address the following questions about the characteristics of streaming media content currently stored on the Web.

— What are the most popular streaming media products used to store freely available audio and video on the Web? Previous research [Li et al. 2002] has shown that proprietary encoded media products utilizing the same network bitrates differ in their impact on streaming network traffic performance. Similar to the situation in 1997 when the large user base for MPEG, AVI, and QuickTime was an obstacle for incoming streaming technologies, quantifying the current dominant technologies used to create streaming media clips can uncover new obstacles for future media applications.

— What is the ratio of streaming audio clips to streaming video clips freely available on the Web? The type of media, whether audio or video, stored on the Web gives researchers indications as to current users' bitrate expectations when streaming over the Internet. Streaming audio often requires only modest bitrates but typically has very discrete encoded bitrate levels. Video, on the other hand, is often bitrate-hungry and can stream over a wide range of encoded bitrates.

— Are the media playout durations stored in media clips long-tailed? Selfsimilar traffic is difficult to manage and there have been a number of studies of Internet traffic patterns that suggest self-similarity (see Park and Willinger [2000] for a survey). Long-tailed distributions of transfer times [Paxson and Floyd 1995;Willinger et al. 1995; Feldmann et al. 1995] may contribute to the self-similarity of Internet traffic. If the distribution of playout durations stored within media clips can be shown to be long-tailed, then this provides evidence to support the conjecture that the distribution of streamed media traffic on the Internet is self-similar.

— What are typical streaming media target bitrates? When encoded, streaming media clips use a target bitrate that has a direct impact on the network traffic rate the media will experience when streamed. Video target bitrates are influenced by such parameters as frame resolution, frame rates, and color depth. Knowledge of stored target bitrates provides insight into the strategies that media content providers use to deal with limited capacities encountered at last-mile connections.

— What fraction of the streaming media codecs available are being used? Innovative compression technologies in new codecs have the potential to deliver higher quality video with lower bitrates. Moreover, new codecs incorporate technologies that yield more sophisticated behaviors that adapt to network conditions to improve quality and performance. Understanding the percentage of older codecs that persist on the Web provides information as to the speed at which new codec technologies are deployed. This article provides detailed information to answer these questions about streaming media stored on the Web today. Since commercial products by their sheer volume have had a strong influence on streaming traffic, our analysis focuses on commercial streaming products such as Microsoft's Media Player, Real Networks' RealPlayer, and Apple QuickTime. Unlike other measurement studies that have tended to view real streaming traffic by monitoring behavior near clients or servers [Chesire et al. 2001; Mena and Heidemann 2000;

Partner websites:

Platinum Television Group -
Americantiredepot - Toyo tires - Offers affordable car tires and wheels in Southern California including Burbank, Glendale, Los Angeles, Long Beach, Van Nuys, Santa Monica, and Whittier.