Automating videoconferencing applications (VCA)

This is a repo containing the code used to automate video conferencing calls in Measuring the Performance and Network Utilization of Popular Video Conferencing Applications.

---

This code allows you to simulate videoconferencing calls on three major applications: Google Meet, Zoom, and Microsoft Teams. This code relies on the python guibot module. We have included images used by guibot to interact with the applications.

Dependencies

We use python to automate these calls and require the following packages:

• guibot
• pyautogui

Automating many calls and limiting available bandwidth

We use static.sh to automate many calls and limit the available bandwidth. It is a wrapper script for test.py, which is explained in the next section. In our experiments, we shaped the interface on the router instead of on the devices. As a result, static.sh is configured to communicate with the router. It is possible to configure static.sh to shape on the device by omitting ssh -n $ROUTER from lines 22 and 25 and deleting line 8. It should be noted that we use tc to shape interfaces, so it may be difficult to shape locally on non-Unix machines.

Assuming static.sh is correctly configured, you can automate calls using the following command:

./static.sh [VCA] [DURATION] [CAPTURE INTERFACE] [SHAPING INTERFACE] [TRACE] [URL]

• VCA is either 'zoom', 'meet', or 'teams'.
• DURATION is the call duration.
• CAPTURE INTERACE is the local interface that we capture network traffic on.
• SHAPING INTERACE is the interface that we shape (either locally or at the router).
• TRACE is the path to a trace file. static.sh will iterate over each line of the trace file and shape the indicated interface accordingly. The format of each line of the trace file is [DURATION] [DOWNLINK BANDWIDTH] [UPLINK BANDWIDTH]. The bandwidths should be given in Mbps. A sample trace file is included in the repo.
• URL is the url of the meeting to join.

Automating a single call

Alternatively, you may use test.py independently of static.sh. test.py will automate one call. View all of the configuration options with python3 test.py -h. There is an option to launch in browser vs. using the client (for now we assume the application executables are in the working directory). You may also choose to capture the network traffic using the -r flag. Note: if you choose to capture traffic you must also specify an interface to capture on using the -i flag. A sample command could be:

python3 test.py [VCA] [DURATION] -i [INTERFACE] -id [URL] -r [DOWN]-[UP]

• VCA is either 'zoom', 'meet', or 'teams'.
• DURATION is the call duration.
• INTERFACE is the capture interface.
• URL is the meeting url
• DOWN-UP are identifiers used for logging

---

Data Collected

Executing the sample command will automate the calls, save network traffic and grab WebRTC (for meet and teams). Network traffic is saved in a directory called captures that is in the working directory. WebRTC stats are saved as jsons in a directory called webrtc in the working directory. Finally, we use a log file to save metadata on the output stats and pcaps. Each pcap and json are given a unique ID. The network parameters, applications used, and whether it was used in browser or client are associated with this unique ID in a file called stats.log, also stored in the working directory.

---

Requesting data

We are happy to share the data we collected from our experiments upon request.