The Need for Speed: Cutting Deployment Times (Again!)

Last year we illustrated how reducing average deployment time could free up people. At that time, deployment times were reduced by about half, and this led to just under twice as many deploys for 2023.

Building on this success, in the past year there's been increased awareness to making deploys faster. In 2024 we can now deploy by two-thirds less time vs 2023 (and in certain cases much more).

2023
2024

There was no special effort, dedicated project or KPI to drive this. Just access to the data and having some awareness of how it might be improved.

Highlights Of Changes:

• Improving sql indexing sped up not just our tests but all users of the application.
• Regularly profiling the most performance-influencing modules in the application again not only reduces deployment times but also helps end users.
  • This resulted in re-implementing some high-impact code to get a big speedup.
  • After, it came up why sql didn't already deliver something like our proof-of-concept-refacotr?
  • Looking for this led to a sql config change that provided an even bigger speedup (vs PoC refactor).
• Making new tests parallelizable by default.
• Having regular discussions about idealized deployment times and next constraints.

In 2024 the reduced deployment times did again lead to a substantial increase in deployments. There were just under 70% more deployments relative to 2023.