Bandwidth Estimation Techniques for Relative ‘Fair’ Sharing in DASH

In the adaptive video streaming (AVS) literature the term fair sharing has been used to describe equal amounts of bandwidth allocated to adaptive client players. However, we argue that even though bandwidth sharing is an important aspect in some problems the same does not apply to AVS. Here the term relative ‘fair' sharing is more applicable. The reason is that videos have different quality levels and will require differing amounts of the bandwidth to satisfy their needs. A 90% to 10% sharing may be sufficient for two players, one with high demands and the other with low demands. A 50% sharing may lead the player with the high bandwidth demand to get too little of the needed bandwidth resource. In addition, channel conditions may lead to players requiring different amount of bandwidth. Again, the concept of fair sharing has to be extended to relative ‘fair' sharing for such scenarios. Hence, bandwidth estimation techniques players use to estimate the network bandwidth is very important in segment selection. A player utilizes one of the many techniques to determine what share of the bandwidth it can utilize among competing players. In this paper we explore some of the techniques used in state of-the-art players in their attempt to obtain a ‘fair' share of the network bandwidth.