Talk

FastAPI & Benchmarks

Thursday, May 29

12:25 - 12:55
RoomTortellini
LanguageEnglish
Audience levelIntermediate
Elevator pitch

Anyone can go and create a benchmark, but can you trust it? šŸ˜…

In this talk, The FastAPI Expert will present you numbers you can trust, and explain the reason for them.

Heā€™ll also explain how you can improve the performance in each situation!

Abstract

Iā€™m not a huge fan of benchmarks, since most of them arenā€™t accurate.

During the last years, Iā€™ve seen a lot of benchmarks, and none of them seem to be fair. To be fair, I donā€™t think information is well spread when the matter is ā€œhow to be more performantā€, so we see a lot of naive benchmarks around.

Nonetheless, when performed correctly, trying to analyze the numbers, we can understand the pitfalls, and improve the applications, or even the frameworks itself. For that reason, this talk will be about benchmarks and getting knowledge about those numbers.

In this talk, weā€™ll see a bit about HTTP parsing, JSON parsing for request and response, files parsing, dependency injection, async def vs def, and much more. The idea is to compare similar solutions, and leave this talk with concrete knowledge about how you can improve your own deployed FastAPI application.

TagsBest Practice, Web Frameworks
Participant

Marcelo Trylesinski

Maintainer of Starlette & Uvicorn, currently working at Pydantic! āœØ