We’ve last benchmarked ara itself as well as different versions of python and ansible but it’s been a while !
Since then:
ansible-core 2.11, 2.12, 2.13 and 2.14 have released (time flies!) python 3.10 as well as 3.11 have come out ara has had a few releases with the latest being 1.6.1 Python 3.11 is reported to be up to 10-60% faster than Python 3.10 and it felt like a good opportunity for some new benchmarking metrics including database backends....
I must preface this blog post with a disclaimer.
I’ve debated whether this should go in documentation and settled on a blog post instead because there are some opinions involved and it’s not something the project can (capital S) Support. I wouldn’t say no if someone wanted to contribute documentation or commit to support it but this has so far not materialized since the project exists.
In other words: consider this is a humble holiday season gift to you in hope that it is useful but with no guarantee....
In the last blog post, we benchmarked the overhead of recording Ansible playbooks with ara.
There’s still room for improvement but we measured that it was on the right track and getting better. The data is useful so we can tell whether we’re improving things or making them worse !
ara 1.5.5 adds support for searching playbooks by Ansible versions and I was curious what kind of data we would get if we ran our benchmark with different versions of Ansible....
We’ve had issues opened about benchmarking and improving the performance of the ara callback plugin as well as the API server for a while now and I only recently took a bit of time to tackle the callback.
It paid off because we can already see significant performance benefits in the 1.5.3 release of ara and it builds a foundation for future improvement opportunities.
If you’d like to see the raw unformatted data that was used for this post, check out this gist on GitHub....