jitsi-load-test-cli

A simple bash-based CLI tool for performing load tests on a Jitsi-Meet instance with customizable parameters. This tool uses a dockerized version of Gst-Meet to simulate participants.

By default the media used for the video publishing is stored in the media folder and it's a small VP8 Big Buck Bunny.

Features

• Specify room and domain for the load test
• Configure the number of video and audio publishers
• Set the number of subscribers
• Define the duration of the load test
• Define the media file to use
• Define the lastN value
• Configure XMPP domain, Focus JID, and MUC domain (with defaults)

Known limitations

• Your target Jitsi infrastructure must use XMPP over WebSocket for Jingle signaling.
• Your target Jitsi infrastructure must have Colibri over WebSocket enabled, as Gst-Meet does not yet support SCTP for Colibri traffic.
• You must use already encoded media files (vp8, vp9 or av1).

Gst-Meet docker image

As no official image of Gst-Meet is provided today you can build your own from the official Gst-Meet repos or you can use my own daimoc/gst-meet.

By default the script uses the daimoc/gst-meet image.

Requirements

• Bash (Unix shell)
• Docker

Installation

1. Clone the repository:

   git clone https://github.com/daimoc/load-test-cli-tool.git

2. Navigate to the repository directory:

   cd load-test-cli-tool

3. Make the script executable:

   chmod +x load_test.sh

4. Install docker

   apt install docker.io

Usage

Run the script with the required options:

./jitsi-load-test-cli.sh --room ROOM_NAME --domain DOMAIN --video-publishers NUM_VIDEO_PUBLISHERS --audio-publishers NUM_AUDIO_PUBLISHERS --subscribers NUM_SUBSCRIBERS --duration DURATION_SECONDS

Example

./jitsi-load-test-cli.sh --room "bob0" --domain "meet.jitsi" --video-publishers 6 --audio-publishers 5 --subscribers 10 --duration 60 --token XXXXXXX

Result view from a browser user :

Options

• --room - Name of the room
• --domain - Your Jitsi domain (e.g. meet.jitsi)
• --video-publishers - Number of video publishers
• --audio-publishers - Number of audio publishers
• --subscribers - Number of subscribers
• --duration - Duration of the test in seconds
• --media - Media file to use for publishers (it should be in webm format)
• --video-codec - Video codec used by video sender agents. It must match your media file because we don't want codec transcoding in a load testing tool (default: vp8)
• --token - JWT token to run test on Jitsi-Meet with authentication enabled
• --last-n - Last-N value setting for subscribers to limit received video streams per subscribers (default: 25)
• --room-numbers - Number of rooms created for the test. Each room is named $room_$index and will have configured the same video-publishers, audio-publishers, and subscribers. By default only one room is created with name = $room
• --xmpp-domain - XMPP domain (default: meet.jitsi)
• --focus-jid - Focus JID (default: focus.meet.jitsi)
• --muc-domain - MUC domain (default: muc.meet.jitsi)

JWT Token Warning

If you want to run your test with a Jitsi-Meet installation with JWT authentication enabled you need to provide a valid token to the test-cli script. And if you need to use the multiple room generation option (--room-numbers), you need to provide a token valid for all the rooms (room:* in the token payload).

Media files

Currently, I don't know if simulcast is possible at the sender part of Gst-Meet. It is still under discussion in the Gst-Meet repository: here.

So, I ran tests with three different media qualities to match WebRTC simulcast used by default Jitsi:

• High: 1280x720, VP8, Target Bitrate 1.5 Mb/s
• Standard: 640x360, VP8, Target Bitrate 500 kb/s
• Low: 320x180, VP8, Target Bitrate 200 kb/s

You can generate your test files with FFmpeg from your favorite source video (Big Buck Bunny, Narwhals...):

ffmpeg -i INPUT_FILE -r 30 -s 1280x720 -c:v libvpx -quality realtime -keyint_min 120 -minrate 1.5M -maxrate 1.5M -b:v 1.5M -c:a libopus -ac 2 result_high.webm 

ffmpeg -i INPUT_FILE -r 30 -s 640x360 -c:v libvpx -quality realtime -keyint_min 120 -minrate 500k -maxrate 500k -b:v 500k -c:a libopus -ac 2 result_standard.webm 

ffmpeg -i INPUT_FILE -r 30 -s 320x180 -c:v libvpx -quality realtime -keyint_min 120 -minrate 200k -maxrate 2M -b:v 200k -c:a libopus -ac 2 result_low.webm 

The default file is bbb.webm, which is a Big Buck Bunny VP8 284x160 200 Kbit and Opus 48000 96 Kbits.

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any changes or improvements.

License

This project is licensed under the Apache-2.0 License - see the LICENSE file for details.

Author

• Damien Fetis daimoc

Acknowledgments

• Inspiration: this tool is inspired by the LiveKit CLI tool