@oaklean/profiler

A library to measure the energy consumption of your javascript/typescript code

Usage

1. Installation

npm add @oaklean/profiler

2. Create a .oaklean config file

The @oaklean/cli can be used to easily setup a .oaklean config file.

1. Install the cli: npm add --save-dev @oaklean/cli
2. Run the init script: oak init
3. 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)

4. 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)

Available Sensor Interfaces

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 the sampleInterval options in the .oaklean config file above.

3. Measure your code

• Option 1 (Code Injection):

  ⚠️ On Windows, this feature is not fully supported yet.

   import { Profiler } from '@oaklean/profiler'
   async function main() {
   	await Profiler.inject("<report-name>") // IMPORTANT: need the await
   	// run the code to measure
   }
   main()
  
   // profiler stops and exports profile when applications stops or gets killed
   // If the resp. exports are enabled the profiler will automatically
   // export the measurements into the output directory (defined via the `.oaklean` config) `<rootDir>/<report-name>/<timestamp>/`

• Option 2 (Code wrapping):

   import { Profiler } from '@oaklean/profiler'
   const profile = new Profiler(true, 'profile-name')
  
   async function main() {
   	await profile.start("<report-name>")
   	// run the code to profile
   	await profile.finish("<report-name>")
   }
   main()
  
   // If the resp. exports are enabled the profiler will automatically
   // export the measurements into the output directory (defined via the `.oaklean` config) `<rootDir>/<outDir>/<report-name>/`

4. 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:

• Github
• VS Code Extension

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: