Evaluating QoS and QoE Dimensions in Adaptive Video Streaming

Abstract

The focus of this thesis has been on Quality of Service (QoS) and Qualityof Experience (QoE) dimensions of adaptive video streaming. By carryingout a literature study reviewing the state of the art on QoS andQoE we have proposed several quality metrics applicable to adaptivevideo streaming, amongst them: initial buffering time, mean duration of arebuffering event, rebuffering frequency, quality transitions and bitrate. Perhapscounterintuitively, other research has found that a higher bitratedoes not always lead to a higher degree of QoE. If one look at bitrate inrelation to quality transitions it has been found that users could prefer astable video stream, with fewer quality transitions, at the cost of an overallhigher bitrate. We have conducted two case studies to see if this isconsidered by today s adaptive video streaming technologies. The casestudies have been performed by means of measurements on the playersof Tv2 Sumo and Comoyo. We have exposed the players to packet lossand observed their behavior by using tools such as Wireshark. Our resultsindicate that neither player take the cost of quality transitions intoaccount in their rate adaptation logic, the players rather strive for a higherquality level. In both cases we have observed a relatively large numberof quality transitions throughout the various sessions. If we were to giveany recommendations to the Over-the-Top (OTT) service providers, wewould advise them to investigate the effects of quality transitions andconsider including a solution for handling potentially negative effects inthe rate adaptation logic of the player.