Measurement of Video Conferencing Libraries using a Test Framework based on Profiling

Video Conferencing systems have become highly popular with the explosion of bandwidth and computing power. This has made high quality video conference possible on embedded hand held devices. This has created an ecosystem of companies developing applications which are powered by video conferencing libraries and a parallel ecosystem of companies who make those video conferencing libraries. An application developer has to design his system in a way that the video conferencing library (which is the backbone of the application) is efficiently integrated and is performing optimally. In order to do that, the developer must be acutely aware of features such as platform compatibility, resource usage, performance boundaries and bottlenecks of the library in order to make an informed decision. Hence the developer needs a test-bench which can evaluate the above mentioned features. This thesis mainly discusses the need to arrive at one such test framework followed by the framework itself. The framework designed in this thesis consists of a network emulator which is combined with a popular network protocol analyser. It profiles the library and monitors variables essential to Quality of Service (QoS) like bandwidth usage, frame rates, frame/packet sizes and audio/video delay under varying network conditions. By observing the change in these variables, the QoS offered by the library can be determined. This information can also be extrapolated to understand the challenges in guaranteeing a minimum quality of experience (QoE). This thesis applies the framework on one chosen library (Library 1) and Google Hangouts and discusses the results. The performance of the library under varying network conditions (bandwidth, packet loss and latency) is observed, enabling Presence Displays BV to extract information about how the library works. Presence Displays BV is building a video conferencing enabled product and uses the framework discussed in this thesis to analyse video conferencing libraries in order to build the application around it. It has been observed that Library 1 favours video quality above other parameters and only compromises video quality to accommodate more users participating in a call. Google Hangouts favours audio, motion content in the video, interactivity and robustness over video quality. This information can be used to configure Library 1 in a way such that it delivers maximum QoE in the available resources.

Subject

video conferencing
measurement analysis
quality of experience

To reference this document use:

http://resolver.tudelft.nl/uuid:6f465acb-40b6-47af-8cf1-f3696862d41f

Part of collection

Student theses

Document type

master thesis

Rights

Files

PDF

thesis.pdf

2.05 MB

Close viewer