benchmw is a set of scripts to make benchmarking MediaWiki and then sharing results easier.
You need a server configured to serve MediaWiki, tuned to whatever you want to benchmark. From a server with ab
installed (like deploy1001), run:
python3 run_benchmarks.py mw1234.eqiad.wmnet label
label
should be an internal identifier that describes the configuration of what you're testing.
Repeat with different labels for whatever else you want to compare to.
Move all the *.dat
files into a folder named data
on your machine and make sure you have gnuplot
installed.
First, create a config.yaml
file based on the sample one provided. Specify the human-readable descriptions for the labels used and then what comparisons should be charted.
python3 filter_and_plot.py data config.yaml
That will clean the data and generate images in data/images/
.