State of Devops 2024 - An Ops Take

The DORA State of Devops report for 2024 is out, and it’s… big. So here’s some notes from an IT Ops expert.

Here’s the report, so you can see it for yourself.

A simple point from early in the executive summary is a point worth thinking about.

Fail Fast, Fail Better

Risk vs Reward

When we release, we try to make sure the release is in good shape. We don’t want a broken site; we don’t want to show customers our mistakes. But how long it takes to rewind a release determines how much damage can be done, how many customers see a mistake.

So the degree of caution around release is based on how confident we are in our testing, and how much damage a mistake can do.

Any release can cause harm, but if you can roll back quickly, any release can be rolled back quickly and the harm is very limited. Therefore, you can place your faith in good automated testing, and if that goes wrong, rapid roll back.

A Virtuous Cycle

The reward for this calculated risk taking is that you can release more frequently - because the lower caution allows for a simpler release process. And if you release more often, you put less into each release.

If you put out smaller releases, then assessing their correctness is simpler, and more likely to be done correctly. This increases the faith in testing and assessing post-release. This reduces caution.

This pushes you towards small, frequent releases, that are easy to verify.

Work Well Worth It

Focussing effort on optimising rollback, both in software design and release systems, pays dividends because it frees you to release quickly and easily. Releasing a little and often is what leads to a high release success rate, high feature velocity, happy engineers, and happy customers.

It might seem like planning for failure is not making you any progress against your goals, but it really, really is. Failure is inevitable, but reducing your fear of it is within your grasp.