From 25 Minutes to 2.5: Okta's Journey to Faster Builds
summary
Okta measures full Maven clean install across a 1,400‑module monolith, adds a Maven Build Scanner and a custom aggregator to Wavefront/Prometheus, and then applies targeted fixes: doubling Maven heap cuts median builds from ~21m to ~9m, switching to native ARM64 Java on Apple Silicon halves times again, patching javac with Java 9’s patch module option saves ~15% versus Java 8, and iterating to make the MAVEN cache extension hit 100% yields some developers with ~2.5m median builds.
AI-written summary · about
chapters (7)
- 0:00Introduction to the monolith and developer productivity team
- 1:06Defining the problem-solving cycle: observation, hypothesis, and experiment
- 9:45Implementing Maven build scanner for data aggregation
- 10:46Doubling Maven memory to reduce garbage collection
- 12:04Migrating build environment from Intel to Apple Silicon
- 14:51Java 11 upgrade and compiler optimization
- 17:58Refactoring build process to implement caching mechanisms
related talks
49mAccelerating Maven Builds: From Snail's Pace to Rocket Speed
37mApache Maven 102: Best Practices
01h 39mWorkshop: Maximize Developer Productivity with Fast and Reliable Gradle and Maven Builds
30mHow we brought down our build times. General optimization tips
40mNail your Gradle build time
54mFrom Laptop Builds to Advanced CI
41mBreaking the build
27mFaster builds on Android: from 2 minutes to 2 seconds