# Serving model via Ray Serve vs FastAPI on ECS

**URL:** <https://discuss.ray.io/t/serving-model-via-ray-serve-vs-fastapi-on-ecs/15536>\
**Category:** Uncategorized\
**Created:** [August 12, 2024, 5:31pm UTC](https://discuss.ray.io/t/serving-model-via-ray-serve-vs-fastapi-on-ecs/15536 "2024-08-12T17:31:45Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Wilfredc](https://avatars.discourse-cdn.com/v4/letter/w/82dd89/32.png) [@Wilfredc](https://discuss.ray.io/u/Wilfredc)\
**Post date:** [August 12, 2024, 5:31pm UTC](https://discuss.ray.io/t/serving-model-via-ray-serve-vs-fastapi-on-ecs/15536/1 "2024-08-12T17:31:45Z")

</div>

We have been evaluating Ray Serve and FastAPI for serving models in AWS ECS. The performance for running a single app in a single ECS node, Ray Serve lags behind FastAPI a lots (in terms of requests per second and response time).

In which use cases, Ray Serve outperforms FastAPI when serving a single or 2 models in a single application?

Thank. you