Art der Arbeit:

Masterarbeit Informatik

Status:

Abgeschlossene Arbeit

Ansprechpartner:

• Christian Moldovan

Gutachter:

Michael Goedicke

Gutachter:

• Prof. Dr. Tobias Hoßfeld

Kurzfassung

Video streaming has become one of the most bandwidth-consuming services on the Internet. As in video streaming the video is played while being downloaded, this process is sensitive to variations in the available bandwidth. Especially outages during the transmission can lead to quality degradations or even so called stalling events, in case the playback stops, because the player ran out of data. Stalling has a massive influence on the Quality of Experience (QoE). Variations in the bandwidth can be compensated with buffering. But with a larger buffer, not only the initial waiting time, which denotes the time until a video starts playing, but also the overall bandwidth consumption rises, because a large quantity of the videos are aborted before they end. In numbers, about 40% of all viewed videos on YouTube were aborted within the first 30 seconds, 20% of the videos were played less than ten seconds [Fin+11]. When a video is aborted, all video in the buffer is lost and can be counted as ‘wasted’. This generates significant costs for both user and service-provider. Transmitting data costs money for the provider of the video service, for server capacity as well as for traffic. But the transmission can also induce costs for the user. On mobile plans, it is often paid for a fixed amount of traffic. More traffic equals higher prices. Additionally, transmitting data on mobile devices consumes energy and therefore reduces battery lifetime. Summing up, a compromise between buffering and user experience has to be found. In this work, we investigate two different video players and their behavior under realistic scenarios while being on the move, in typical commuting situations. Hereby, we focus on the effect of buffering on the playback. As the user experience is still a heavily discussed field, we obtain objective Quality of Service key numbers of the playback. These are then be used to derive QoE numbers using different methods.