Where are the Sweet Spots? A Systematic Approach to Reproducible DASH Player Comparisons

by Denny Stohr, Alexander Frömmgen, Amr Rizk, Michael Zink, Ralf Steinmetz and Wolfgang Effelsberg

Dynamic Adaptive Streaming over HTTP (DASH) aims to constantly provide high user quality of experience in dynamically changing network environments. The heterogeneity of the streaming environment makes many of the developed DASH algorithms possess performance affinities that we denote as sweet spots.

We show the substantial impact of the video player choice and its configuration on the streaming performance. We systematically examine three established open-source DASH players, i.e., DASH.JS, Google’s Shaka Player, and AStream, that implement fundamentally different configurations featuring various adaptation algorithms.

We establish a large scale emulation framework to (i) extract player sweet spots and (ii) achieve a direct, reproducible comparison of real-world DASH players and algorithms. In the following we show empirical evidence (exerpts from the paper) demonstrating that an isolated analysis of DASH player modules is insufficient to capture the player streaming performance.


Stalling duration (lower is better) and stalling QoE (higher is better) of DASH Players with different adaptation algorithms, available bandwidth volatilities and segment length configurations. The player and the segment length have a strong impact on the stalling QoE whereas the impact of the adaptation algorithm is marginal.

One of the major observations in this paper is that the choice of the adaptation algorithm is dominated by the choice of the player and its configuration.

Streaming Quality

Mean playback bitrates given fixed available bandwidth and default buffer settings. DASH.JS and Shaka Player show consistent behavior while the AStream emulator shows highly sensitive performance with respect to the configuration and the network environment. The black stairs show the video representations available in the data set. The depicted player profile indicates (i) QoE degradation given networking conditions and (ii) player efficiency, i.e., how much available bandwidth is required to sustain a quality bitrate.

Trade-offs - A Pareto Frontier

Trade-off between adaptation count and stalling QoE for different configurations, aggregated for all analyzed environment conditions.

All players and adaptation algorithms are represented on the Pareto frontier such that no single player / configuration dominates.

Research Paper

The paper “Where are the sweet spots? A systematic approach to reproducible DASH Player comparisons” is accepted as a full research paper at ACM Multimedia 2017.

PDF Bibtex


To enable other researchers to reproduce and build on our work, we provide our evaluation framework.

  1. Setup a new DASH simulation study

  2. Analyse your DASH simulation study


Can be downloaded here

Install under Linux

Extract the source, e.g.,

tar -zxvf acm-maci-dash.tar.gz

and follow the instructions in acm-maci-dash/README.md.

Install under Windows

  • Extract the source
  • Install Docker (e.g., following these instructions).


The MACI Frontend:

  • Attached (logs visible) docker-compose up --build maci-backend

  • Detached docker-compose up -d --build maci-backend

The MACI Worker:

  • vagrant up --provision


This work has been funded by MAKI.


Feel free to contact Denny Stohr and Alexander Frömmgen for any comments and questions.