Flow-level traffic model for adaptive streaming services in mobile networks

Abstract

This paper proposes a traffic model for mobile networks carrying adaptive streaming. The proposed model takes into account the flow dynamics and some application-level parameters influencing the performance of adaptive streaming, such as the presence of a playout buffer, chunk duration configuration and the video bit rate configuration. Paper shows based on flow-level queueing model how to compute performance metrics for a streaming user such as average video bit rate, average service time, average deficit rate, defined as the probability of having its instantaneous throughput lower than the chosen video bit rate and average buffer surplus, related to the amount of video accumulated in the buffer. Since heterogeneous radio conditions and the coexistence of elastic traffic and streaming services make the exact solution intractable, we propose a simple yet accurate approximation that reduces the computational complexity and makes system dimensioning easier for mobile network operators. Our numerical results show the performance impacts of video chunk duration and amount of available video bit rate. These performance impacts obtained by our models give an insight for operators to properly dimension their adaptive streaming service by well configuring both HTTP-level and MAC-level parameters.