Skip to content

Client case study: wikando

This a testimonial from Timo Schäfer, engineer at Wikando, dated March 13 2024.


RunsOn has been amazing to use so far, your service has made massive progress over only these few weeks we’ve been testing it, and those few months since your project launched.


Our original goal was cutting our GH Actions spending, at least below our budgeted limit of $300/mo, which has recently ballooned closer to $600/mo.

Before looking into alternatives, we were using ubuntu-latest and ubuntu-latest-4-cores GH hosted runners, but hated their performance - even the larger size performed worse than late 201X Intel i5 MacBook Pros running our dockerized testing. The cost of even larger runners was deemed not viable for the performance we were getting out of those smaller instances.

We evaluated self-hosting an existing GH Action Controller implementations for ephemeral runners - and even just migrating to GitLab - which is when I stumbled upon RunsOn and pitched it for evaluation, thanks to your licensing terms.

RunsOn allowed us to finally utilize our knowledge of our CI and testing tools - which ones benefit from more VCPUs and which ones will run fine on even single CPU burstable instances - to not only save on costs (our original goal), but also massively on CI runtime.


Without spot and burstable instances included in the calculation we were already expecting our costs to drop by ~75% (+- some due to our decision to put RunsOn in an isolated AWS account):

  • 21453 minutes 2CPU, 24455 minutes 4CPU
    ~$563 with GH hosted runners, ~$136 for compute on RunsOn.

  • our main repository saw an average decrease in CI runtime of 8 minutes (23 -> 15).

  • another repository could fully utilize 16 threads (diminishing returns after, so stuck with that number) for one of its heavier analyses, which dropped its runtime by 15 minutes (20 -> 5).

The monetary savings are huge, don’t get me wrong - but those mainly make our finance department happy.

I’m wholly convinced RunsOn will help us deliver software to our customers faster by reducing software cycle times (mostly code reviews).