https://github.com/piccolo-orm/asgi_server_performance
Load testing the different ASGI servers to see which is fastest
https://github.com/piccolo-orm/asgi_server_performance
asgi locust
Last synced: 8 days ago
JSON representation
Load testing the different ASGI servers to see which is fastest
- Host: GitHub
- URL: https://github.com/piccolo-orm/asgi_server_performance
- Owner: piccolo-orm
- Created: 2021-02-28T10:04:12.000Z (about 4 years ago)
- Default Branch: master
- Last Pushed: 2022-04-14T05:53:27.000Z (about 3 years ago)
- Last Synced: 2025-03-28T19:44:42.990Z (26 days ago)
- Topics: asgi, locust
- Language: Python
- Homepage:
- Size: 673 KB
- Stars: 9
- Watchers: 2
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# ASGI Server Performance
Comparing the performance of various ASGI servers.
## Setup
Install the requirements:
```bash
pip install -r requirements.txt
```
Start one of the ASGI servers, using the named shell script. For example:
```bash
# Increase the max number of files which can be open:
ulimit -Sn 10000
# Start the server:
./daphne.sh
```
Start Locust, and three workers, each in a separate terminal tab.
```bash
# tab 1
locust --master
# tab 2
locust --worker
# tab 3
locust --worker
# tab 4
locust --worker
```
Open `http://0.0.0.0:8089/` in a browser to see the Locust GUI. Start a new test.
### Number of Locust workers
The test was run using a 2.7 GHz Core i7 (9th Gen) processor with 6 cores, and 16 GB RAM. Your mileage may vary - make sure there are enough workers that none of them hit 100% CPU usage, as it will give misleading results.
You can see the CPU usage of each worker in the workers tab of the Locust GUI.
### Server configuration
To get fair results, logging was disabled for each server.
-----
## Results
Uvicorn has the largest throughtput, and Hypercorn and Daphne were basically equal.
### Uvicorn
### Hypercorn
### Daphne
-----
## Conclusions
If performance is your main aim, then Uvicorn is a good starting point. However, all of the servers performed well, so all are good choices.