Energy consumption of videoconferencing apps

In this project we, the Super Green Lab team, want to do research on the energy consumption of the two popular Android videoconferencing apps Zoom and Google Meet. This repository is devided into two main parts: the actual experiment (android-runner) and the analysis of the measured data (data-analysis).

android-runner

Python framework for automatically executing measurement-based experiments on native and web apps running on Android devices

Data analysis

The data analysis can again be split up into two parts: the preprocessing and the actual analysis.

Preprocessing

In order to be able to evaluate the measured data, it is reasonable to extract the relevant information that Batterystats has written into multiple files spread over multiple folders. We used a Python script to do this. Before running, a few parameters have to be adapted in this script:

• phone: The ID of the used phone as specified in Monkeyrunner (e.g. motog5)
• raw_data_path: The path to the folder containing the raw data generated by the experiment.

There are two preprocessing scripts available: one for the preliminary experiment (preprocessing-pre.py) and one for the main experiment (preprocessing-main.py). They slightly differ, since the folder structure and the interesting attributes also differ. Both are generating a CSV file called results-pre.csv or results-main.csv respectively in the raw data’s parent folder. This CSV file contains the following attributes:

Attribute	Description	Values
app	The examined videoconferencing app	meet or zoom
number_of_participants (only in main)	Number of participants in the video conference	2 or 5
camera (only in main)	Activation state of the camera	True or False
microphone (only in main)	Activation state of the microphone	True or False
background (only in main)	Activation state of the virtual background	True or False
duration_in_minutes (only in preliminary)	Duration of the video conference in minutes	Numerical value
joules	Consumed energy in Joules	Numerical value

Folder structure

Inside the specified raw_data_path folder must be the following folder structure for the preprocessing script to find the measurements and used treatments.

Preliminary experiment:

• Folders with arbitrary names, all containing:
  • Config file: config.json
  • Folder: data
    • Folder: Name of the phone (as specified in the variable phone)
      • Folder: Either com-google-android-apps-meetings or us-zoom-videomeetings
        • Folder: batterystats
          • CSV files starting with Joule containing the energy measurements

Main experiment:

• Folders with arbitrary names ending with either -2p (for two participants) or -5p (for five participants), all containing:
  • Folders with arbitrary names ending with a hypen (-) followed by a subset of the letters c (iff the camera was active), b (iff the virtual background was active) and m (iff the microphone was active) – for example: somefoldername-cbm
    • Config file: config.json
    • Folder: data
      • Folder: Name of the phone (as specified in the variable phone)
        • Folder: Either com-google-android-apps-meetings or us-zoom-videomeetings
          • Folder: batterystats
            • CSV files starting with Joule containing the energy measurements

Acutal analysis

To analyse and visualize the preprocessed data, we used R. With R it is very simple to read CSV files and perform analyses on the read data. For creating plots, we used ggplot2.