Choosing the Level of Stream Adaptation in TCP-Friendly Congestion Controlled Video Streaming Servers

In this paper we investigate stream adaptation policies for video streaming servers which use smooth TCP-friendly congestion controls to compute the transmission rate. TCP-friendly congestion controls compute, by specific means, the rate at which the streaming server can send video data (i.e. available bandwidth in the network). Of course, this rate must be supported at application level by an appropriate stream adaptation mechanism. We seek an optimal adaptation policy which maximizes the perceived quality at the client side and tries to avoid an empty prefetch buffer at the client side, in the given TCP-friendly conditions. We formulate our problem in an optimization framework and then devise a gradient ascend algorithm to solve it. The solution is an adaptation policy that can be used to support TCP-friendly rate control at codec (application) level.