TechPrufer
Get started
Meta
Meta logoMeta · Feature trail

Serverless Jupyter Notebooks

1 post · Jun 2024 – Sep 2024

What this trail covers

Meta's Bento notebook platform has introduced serverless execution for 'lite' workloads by leveraging Pyodide and WebAssembly. This allows code execution directly in the browser, eliminating the need for remote compute provisioning for simpler tasks. The system integrates with existing cell magics and uses web workers to prevent browser lock-up. Future plans include improving workload heuristics, defaulting to serverless, and exploring hybrid execution for non-lite cells.

All Meta trails
  1. Inside Bento: Jupyter Notebooks at MetaThis post details the integration of Pyodide and WebAssembly into Meta's Bento notebook platform to enable serverless execution of 'lite' workloads directly in the browser. This eliminates the need for remote server components for certain tasks, improving efficiency and accessibility. The system leverages web workers to maintain browser responsiveness and plans to explore hybrid execution models for more complex workloads.
  2. Serverless Jupyter Notebooks at MetaIntroduced serverless execution for 'lite' Jupyter notebook workloads within Meta's Bento platform by integrating Pyodide and WebAssembly. This enables code execution directly in the browser, removing the need for remote compute provisioning for these workloads. Implemented a kernel abstraction to support both traditional server-based kernels and the new browser-based serverless kernel seamlessly. Developed middleware to capture and process existing cell magics (e.g., %%sql, %%googlesheet, %%graphql) in JavaScript, injecting results back into the Python kernel. Utilized web workers to prevent browser lock-up during 'expensive' kernel operations. Implemented data upload functionality using an async job with Tupperware for non-lite cells.