JupyterLab Benchmarks Documentation

Benchmarking tools for JupyterLab

This project is a place to make JupyterLab faster. For now, we want to optimize for various notebook content:

1. Opening a new notebook.

2. Switching tabs. An acceptable time is 300-500 ms switch time.

3. Closing a notebook

The following is out-of-scope, but could become later in-scope:

1. Initial server start.

2. Initial HTML page loading.

3. Notebok saveing time.

4. Build time.

We iterate in steps:

1. Identify the types of issue and measure what is going on with e.g. experiments and profiling. We also understand the notebook lifecycle

2. Build benchmark tooling to have a baseline and be able to compare with fixes.

3. Define fix strategies and implement those fixes.

Site Contents

Issues

• Issue Types
  • Issue Type 1 - Many Cell Editors
  • Issue Type 2 - Large Text Output
  • Issue Type 3 - Many Rich Output
  • Issue Type 4 - Large .ipynb file
• User Reports

Measures

• Experiments
  • Simple Notebook
  • Editor Comparison
  • CodeMirror
  • Lumino actions on resize
• Lifecycle
  • Small Notebook Lifecycle
  • Medium Notebook Lifecycle
• Browser Profiling
  • Open Many Cells
  • Open Single Cell with many DIV outputs
• Jupyter Telemetry
• Jaeger
• Lighthouse
  • Using throttling
  • Comparing results

Benchmarks

• Benchmarks Tooling
  • Running test
  • Understanding the tests
  • Test notebooks
  • Requirements
  • Customization
  • Benchmarks comparison
• Benchmarks on CI

Memory Leaks

• Memory Leak Tooling
  • Running test
  • Finding leaks
• Memory leak on CI

Fixes

• Fixes Strategies
  • Strip Output
  • Block Div
  • Virtualized Notebook
  • Shadow DOM
  • Tune CodeMirror Configuration
  • Enhance CodeMirror Code Base
  • Content Visibility
  • Fast DOM
  • Web Workers
  • Service Workers
  • Use requestAnimationFrame (non concluding)
  • Update Editor on Show
  • Use pushAll cells insted of push (non concluding)
  • React Concurrency
  • Reuse contentFactory
  • Use scrollbarStyle: ‘null’ in Editor Config
  • Transfer Content in Chunks
  • Server Memory cache
  • Improving Network Performance
  • Web Render
• Fixes Implementations
  • Virtual Notebook Implementation
  • Content Visibility Implementation
  • Block Div Implementation
  • Server Parsing Implementation
  • Shadow DOM Implementation
  • StrippedOutput Implementation
  • VirtualOutput Implementation

Links

• GitHub repo