A jest environment to collect energy measurements on source code level during test execution.
npm add --save-dev @oaklean/profiler-jest-environment
Adjust the jest config like this:
module.exports = {
...
testEnvironment: '@oaklean/profiler-jest-environment/env.js', // add this line
globalSetup: '@oaklean/profiler-jest-environment/setup.js', // add this line
globalTeardown: '@oaklean/profiler-jest-environment/teardown.js', // add this line
...
}
The @oaklean/cli
can be used to easily setup a .oaklean
config file.
- Install the cli:
npm add --save-dev @oaklean/cli
- Run the init script:
oak init
- It will ask you which sensor interface should be used for energy measurements:
Select a sensor interface (recommended for your platform: perf)
None (pure cpu time measurements)
powermetrics (macOS only)
❯ perf (Linux only)
energy measurements on Linux (Intel & AMD CPUs only)
- The cli asks you to confirm your choice and generates a valid
.oaklean
config file for you:
? Select a sensor interface (recommended for your platform: perf) perf (Linux only)
{
"exportOptions": {
"outDir": "profiles/",
"outHistoryDir": "profiles_history/",
"rootDir": "./",
"exportV8Profile": false,
"exportReport": true,
"exportSensorInterfaceData": false
},
"projectOptions": {
"identifier": "XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX"
},
"runtimeOptions": {
"seeds": {},
"v8": {
"cpu": {
"sampleInterval": 1
}
},
"sensorInterface": {
"type": "perf",
"options": {
"outputFilePath": "energy-measurements.txt",
"sampleInterval": 100
}
}
}
}
? Is this OK? (yes) (Y/n)
SensorInterface | Operating System |
---|---|
powermetrics | macOS |
perf | linux |
If you want to how to setup the Sensor Interfaces and how to make them work with Docker you can read more about it here
⚠️ Most Sensor Interfaces need root privileges
Look into the Sensor Interface Docs to see how you can run them without root privileges
🔍 How measurements work
During the test execution measurements are collected with a sample based approach. So for every n - microseconds it collects a v8 cpu profile and energy measurements of the sensor interface. You can adjust the sampling rate with thesampleInterval
options in the.oaklean
config file above.
⚠️ The measurement data is NOT automatically collected
To enable measurements during test execution you must set an environment variable beforehand (ENABLE_MEASUREMENTS=1). When running Jest tests with this environment variable, the measurement data should be saved in the specified output directory.
🤌 For more accurate measurement results
It is recommended to run Jest with the --runInBand option. This ensures that the tests are executed sequentially and not in different threads: npx jest --coverage --runInBand.
Execute the tests:
ENABLE_MEASUREMENTS=1 npx jest --coverage --runInBand
5. Interpret the measurements and determine the source code locations with the most energy consumption
The Oaklean
VSCode Extension lets you to interpret the measurements. It integrates the energy measurements directly into your IDE.
You can find it here:
It provides code highlighting to point out which source code locations consume the most energy:
It also provides multiple features to determine the components that consume the most energy, including node modules: