Analysis of web-based video delivery

The recent shift from specialized video streaming solutions to user-friendly Web-based video delivery platforms has attracted large audiences to these platforms. However, this could also result in future bandwidth bottlenecks and a degradation of network parameters if traffic demands continue to rise. Due to the reliable TCP protocol the influence of network degradations expresses itself only in the playback and buffering behavior of the media player but not in a loss of video quality. In this paper, we describe and conduct performance measurements of the YouTube platform as the largest representative of Web video delivery platforms. First we give insights into the composition of its architecture. Then we describe the video streaming process for which we present models of the client's approaches to video stream buffering and playback control. Afterwards the performance of Web streaming is evaluated using this model in a network emulator. We find that YouTube's rate limiting provides sufficient QoE to the user during good network conditions, but degrades fast with increased packet loss and latency.