goodoldneon/framework-load-test

Server Load Test

Simple (and probably reductive) repo for load testing HTTP server frameworks.

Installation

Run docker build . -t server-load-test to build the server Docker image.

Usage

1. Start the server you want to load test:

   # Express
   $ docker run -i -p=0.0.0.0:8080:8080 --cpus=1 server-load-test node server/express_server.js
   
   # Flask
   $ docker run -i -p=0.0.0.0:8080:8080 --cpus=1 server-load-test python server/flask_server.py
   
   # Flask with Waitress
   $ docker run -i -p=0.0.0.0:8080:8080 --cpus=1 server-load-test python server/flask_waitress_server.py
   
   # Gin
   $ docker run -i -p=0.0.0.0:8080:8080 --cpus=1 server-load-test server/gin_server
   
   # Tornado
   $ docker run -i -p=0.0.0.0:8080:8080 --cpus=1 server-load-test python tornado_server.py

2. Run load-tester/main.ts in a loadimpact/k6 Docker container. Some example usage:

   # 1000 requests
   docker run -i loadimpact/k6 run - <load-tester/main.ts -i 1000
   
   # 1000 requests split between 2 users
   docker run -i loadimpact/k6 run - <load-tester/main.ts -i 1000 -u 2
   
   # View K6 help
   docker run -i loadimpact/k6 run -h

Results

The following table shows requests/second for 10k requests. The tests were run on a 2.6 Ghz i7 MacBook Pro.

Framework	1 User	5 Users	10 Users
Express	620	1970	2520
Flask	270	Failed	Failed
Flask w/ Waitress	500	1070	1090
Gin	590	1620	2230
Tornado	560	1570	1790