This is a fork of https://github.com/klen/py-frameworks-bench, refactored a bit to enable different kinds of experiments.
The tests themselves are orchestrated using Make. In order to run these demos, you will require the following tools (left as an excersize for the reader):
- wrk (https://github.com/wg/wrk)
- docker (https://www.docker.com/)
- virtualenv
- python3.6 (the benchmarks should be run with this version of Python)
The execution is wrk benchmarking an application, sending 400 concurrent requests over 10 threads. The type of work depends on the type parameter passed to make, but is typically some remote call: see the Makefile for more information.
The benchmarks themselves are stored in makefiles under the benchmarks directory.
First, start by making your lab, if you haven't already. This starts up the docker container to execute remote calls against.
make lab
For example, running async-vs-threads would be:
make -f benchmarks/async-vs-threaded.makefile main
The benchmark files themselves are very barebones, just running the main benchmark file with different configuration.
Results are written to results.csv, regardless of the benchmark. You will have to clear this file in between benchmark runs.
The columns are declared in scripts/cvs-report.lua
- Amber Brown (https://github.com/hawkowl)
- Giovanni Barillari (https://github.com/gi0baro)
- Kirill Klenov (https://github.com/klen)
- Yusuke Tsutsumi (https://github.com/toumorokoshi)