If you've ever run sysbench, you know it's impossible to reasonably compare more than 2 tests at once. Beyond that, you need graphs. Seems like everyone has their own home-grown set of little scripts to parse sysbench's output and turn it into a CSV or something else. This is my attempt at it. I tried various existing things but none of them satisfied me.

HOWTO USE

When you use sysbench, you're typically trying to find which configuration works best among a set of different configs. A "configuration" could be a certain type of HDD vs another one, a filesystem vs another one, etc. For a given configuration, you typically run a range of tests, with different parameters (e.g. different number of concurrent threads and whatnot). This script simply wants you to put the output of all the tests you ran for a given configuration into one directory. Look at the sample script included in this repository, runtests.sh, for an example.

So using the example included in this repository: mkdir ~/out cd ~/out ~/runtests.sh ~/sysbench-log2json.py /

Now you should have a big results.js', which you need to put in the same directory as sysbench.html'. Open the HTML file and enjoy pretty graphs.