TyStream

This is a final year project in City University of Hong Kong. Project code: 19CS038.

Contact: [email protected]

Abstract

Video streaming is one of the most popular services on the Internet. Researchers have endeavored to develop new algorithms and protocols across the application layer and transport layer to improve the performance of such service. We propose TyStream, a generic evaluation platform, to simplify the evaluation process of these algorithms and protocols at different computer network layers. We define four network parameters and decouple them in our design. Results show that our evaluation platform can test the supported algorithms correctly and the design logic of TyStream ensures that this system is convenient, efficient, and extensible.

Installation

Tested on: Ubuntu 18.04

• Mahimahi: Build from source by referring to the official guide. Then replace packet/packetshell.* with our files in mahimahi

• Video servers:

  • Apache: Ubuntu built-in. Copy everything inside www.quictest.com to /var/www/html. Delete the HTML header in each file.
  • QUIC: Read Quic Server Installation Guide

• Both Python2 and Python3. Recommend managing environments with conda. Packages:

  • numpy
  • matplotlib
  • pyautogui

• Google chrome browser

Usage

CONFIG

Four sets of configurations are supported: trace, transport, cc, abr. A list of supported parameters are in config/supported.json

Check current configuration:

config

Switch a configuration:

config config_key config value

For example:

config trace ATT-LTE-driving

EXP

exp run_time

run_time is the number of times that you want to repeat your experiment. Must be a positive integer.

Results are stored in exp/results/

PLOT

The visualization configuration consists of external and internal. The internal one will be set to the same as the experiment configuration of the last experiment. If you want to change the default, find it in vis/generic_visualizer.py and rewrite your desired one.

The external one consists of the same set of config_key. 3 of them are fixed and one should be a list. There's another field called dir for the external configuration, which decides the directories of the varied parameter. For example, if your varied parameter is abr with ['mpc', 'robustmpc'], dir1 and dir2, where dir: [dir1, dir2], should contain the log files of mpc and robustmpc respectively.

To change the external configuration:

plot dir n dir1 dir2 ... dirn, or
plot var config_key n config_value1 ... config_valuen

For the above example, it should be

plot dir 2 dir1 dir2
plot var abr 2 mpc fastmpc

To visualizer the results, use the above setting and

plot
plot link_utilization
plot qoe
plot bitrate_selection
plot bandwidth_estimation

, where the default diagram of plot is link utilization. We currently support 4 types of visualization methods.

Acknowledgement

Dash.js and video files in www.quictest.com , and decision servers inabr_servers/are provided by Pensieve with a few modifications. We also learn from Pensieve's evaluation code when designing the work flow of this project.