Docker-Based Evaluation Framework for
Video Streaming QoE in Broadband Networks
Cise Midoglu1
, Anatoliy Zabrovskiy2
, Özgü Alay3
, Daniel Hölbling-Inzko4
, Carsten
Griwodz5
, Christian Timmerer2,4
1
Simula Research Laboratory, 2
University of Klagenfurt, 3
Simula Metropolitan Center for Digital
Engineering, 4
Bitmovin Inc, 5
University of Oslo
OS-05
Motivation
● Video streaming one of the top traffic contributors in the Internet
● Mobile broadband (MBB) networks becoming more and more prominent
● Metadata-rich measurement result
collection is challenging
● Over-the-top (OTT) video analytics
are not used widely in research
OS-05 2
Overview
● Automatised measurement and evaluation framework for video streaming
Quality of Experience (QoE)
OS-05 3
Overview
● Automatised measurement and evaluation framework for video streaming
Quality of Experience (QoE)
○ End-to-end with client and server side
○ Client-initiated active measurements
○ Docker virtualization → platform compatibility
○ Headless streaming → focus on objective QoE
○ Allows to monitor and collect data from multiple layers (application, transport,
network, physical)
○ Allows to benchmark video players, ABR algorithms, networks, operators, ...
OS-05 4
Client Side
● Designed to emulate real user initiating
a video streaming session
● Selenium instrumentation opens
headless browser and streams from
HTTP server
● Implemented as Docker container
● Compatible with mobile broadband
testbeds (e.g., MONROE)
● Configurable (network, player, ABR
algorithm, run duration, batch
experimentation, ...)
● Additional network monitoring (ping +
traceroute), and metadata if available
OS-05 5
Server Side
OS-05 6
● Server hosts landing pages with 3
different video players
○ Bitmovin Player v8
○ Shaka Player v2.5
○ DASH.js Player v2.9
Server Side
OS-05 7
● All test pages use the same asset
○ BigBuckBunny video
○ Encoded in 15 qualities (100 Kbps to
15 Mbps)
○ Retrieved from CDN
Server Side
OS-05 8
● All test pages are integrated with
Bitmovin Analytics web collector v2.3
○ JavaScript executed on client browser
in runtime
Server Side
OS-05 9
● All pages are integrated with Bitmovin
Analytics web collector v2.3
○ More than 70 parameters recorded for
each session (including startup time,
video quality, num. switches, stalls, ...)
Server Side
OS-05 10
● All pages are integrated with Bitmovin
Analytics web collector v2.3
○ Possible to view metrics in real-time
from the Bitmovin Dashboard, or export
in raw format using the Bitmovin
Analytics Export Service
MONROE Integration
OS-05 11
MONROE Integration
OS-05 12
MONROE Integration
OS-05 13
Use Cases and Intended Audience
OS-05 14
Demo
OS-05 15
16
• Two grand challenges
– Improving open-source HEVC encoding
– Low-latency live streaming
• Focus areas in 2020
– Machine learning and statistical modeling for video streaming
– Volumetric media: from capture to consumption
– Fake media and tools for preventing illegal broadcasts
• A workshop (posters/demos) dedicated to middle and high-school students
• Two confirmed keynotes from Google and MIT
• Expecting reduced registration fees thanks to strong support
Important Dates Submit by
Research Track Jan. 10 (firm)
Demo Track Feb. 29
Open Source/Dataset Feb. 29
Workshops Mar. 27
Conference June 8-11
NEW
Visit http://acmmmsys.org today!
NEW
VBIM: Video QoE Dataset Generation
Docker-Based Evaluation Framework for Video Streaming QoE in Broadband Networks
Cise Midoglu, Anatoliy Zabrovskiy, Özgü Alay,
Daniel Hölbling-Inzko, Carsten Griwodz, Christian Timmerer
Simula Research Laboratory, University of Klagenfurt,
University of Oslo, Bitmovin Inc
Open Source Software Competition
Thursday 24 October
10:30 - 12:00
Risso 6

Docker-Based Evaluation Framework for Video Streaming QoE in Broadband Networks

  • 1.
    Docker-Based Evaluation Frameworkfor Video Streaming QoE in Broadband Networks Cise Midoglu1 , Anatoliy Zabrovskiy2 , Özgü Alay3 , Daniel Hölbling-Inzko4 , Carsten Griwodz5 , Christian Timmerer2,4 1 Simula Research Laboratory, 2 University of Klagenfurt, 3 Simula Metropolitan Center for Digital Engineering, 4 Bitmovin Inc, 5 University of Oslo OS-05
  • 2.
    Motivation ● Video streamingone of the top traffic contributors in the Internet ● Mobile broadband (MBB) networks becoming more and more prominent ● Metadata-rich measurement result collection is challenging ● Over-the-top (OTT) video analytics are not used widely in research OS-05 2
  • 3.
    Overview ● Automatised measurementand evaluation framework for video streaming Quality of Experience (QoE) OS-05 3
  • 4.
    Overview ● Automatised measurementand evaluation framework for video streaming Quality of Experience (QoE) ○ End-to-end with client and server side ○ Client-initiated active measurements ○ Docker virtualization → platform compatibility ○ Headless streaming → focus on objective QoE ○ Allows to monitor and collect data from multiple layers (application, transport, network, physical) ○ Allows to benchmark video players, ABR algorithms, networks, operators, ... OS-05 4
  • 5.
    Client Side ● Designedto emulate real user initiating a video streaming session ● Selenium instrumentation opens headless browser and streams from HTTP server ● Implemented as Docker container ● Compatible with mobile broadband testbeds (e.g., MONROE) ● Configurable (network, player, ABR algorithm, run duration, batch experimentation, ...) ● Additional network monitoring (ping + traceroute), and metadata if available OS-05 5
  • 6.
    Server Side OS-05 6 ●Server hosts landing pages with 3 different video players ○ Bitmovin Player v8 ○ Shaka Player v2.5 ○ DASH.js Player v2.9
  • 7.
    Server Side OS-05 7 ●All test pages use the same asset ○ BigBuckBunny video ○ Encoded in 15 qualities (100 Kbps to 15 Mbps) ○ Retrieved from CDN
  • 8.
    Server Side OS-05 8 ●All test pages are integrated with Bitmovin Analytics web collector v2.3 ○ JavaScript executed on client browser in runtime
  • 9.
    Server Side OS-05 9 ●All pages are integrated with Bitmovin Analytics web collector v2.3 ○ More than 70 parameters recorded for each session (including startup time, video quality, num. switches, stalls, ...)
  • 10.
    Server Side OS-05 10 ●All pages are integrated with Bitmovin Analytics web collector v2.3 ○ Possible to view metrics in real-time from the Bitmovin Dashboard, or export in raw format using the Bitmovin Analytics Export Service
  • 11.
  • 12.
  • 13.
  • 14.
    Use Cases andIntended Audience OS-05 14
  • 15.
  • 16.
    16 • Two grandchallenges – Improving open-source HEVC encoding – Low-latency live streaming • Focus areas in 2020 – Machine learning and statistical modeling for video streaming – Volumetric media: from capture to consumption – Fake media and tools for preventing illegal broadcasts • A workshop (posters/demos) dedicated to middle and high-school students • Two confirmed keynotes from Google and MIT • Expecting reduced registration fees thanks to strong support Important Dates Submit by Research Track Jan. 10 (firm) Demo Track Feb. 29 Open Source/Dataset Feb. 29 Workshops Mar. 27 Conference June 8-11 NEW Visit http://acmmmsys.org today! NEW
  • 17.
    VBIM: Video QoEDataset Generation Docker-Based Evaluation Framework for Video Streaming QoE in Broadband Networks Cise Midoglu, Anatoliy Zabrovskiy, Özgü Alay, Daniel Hölbling-Inzko, Carsten Griwodz, Christian Timmerer Simula Research Laboratory, University of Klagenfurt, University of Oslo, Bitmovin Inc Open Source Software Competition Thursday 24 October 10:30 - 12:00 Risso 6