Skip to content

ag-forecasting-api.md

Latest commit

 

History

History
37 lines (27 loc) · 1.63 KB

ag-forecasting-api.md

File metadata and controls

37 lines (27 loc) · 1.63 KB
myst
html_meta
description keywords
A FastAPI/ASGI application used as a case study for porting real workloads to OpenLambda.
OpenLambda, serverless, FastAPI, ASGI, case study

Agricultural Forecasting API

A FastAPI-based ASGI service from the UW–Madison Data Science Institute, used here as a case study for OpenLambda — an existing, real-world application that was never designed for serverless, ported to the platform to find out what OpenLambda needs.

This page focuses on what the port taught us about OpenLambda. For the application itself,
see its [source repository](https://github.com/UW-Madison-DSI/ag_forecasting_api).

Why this application

It's a good stress test for OpenLambda because it is non-trivial and ordinary: a FastAPI (ASGI) backend with real package dependencies, weather-data caching, and parallel compute — the kind of app teams actually run, not a serverless demo.

What it drove in OpenLambda

Porting it surfaced five concrete challenges that drove four new OpenLambda features:

  • Per-function environment variables — to redirect writes away from read-only directories.
  • pip-compile as a lambda — to pin dependencies inside the OpenLambda environment.
  • Direct GitHub deployment — ol admin install straight from a repo URL.
  • Built-in ASGI support — OpenLambda acts as the ASGI server for FastAPI/Starlette/etc.

It also pointed to a future feature: /dev/shm support so process pools work under OpenLambda.

The full porting write-up is in the blog case study.