Speaker: Paul Bakker
See the live blog table of contents
General
- Gave this talk last year, but it changes over time
- Architecture is same as last year. I blogged about this last year
- Java still best radeoff between runtime performance, developer productivity and maintainability
What’s not Java
- JavaScript, Android, Swift for mobile.
- Go for Platform side cars
- Python for “data science”
Spring Boot
- Spring Boot Netflix customization
- Open sourced
- Includes annotation, post processors, etc
- Use SpringBootTest
- Fully on Spring Boot 3 now. Included JavaEE to Jakarta EE and Gradle upgrade
- Starting migration to Spring Boot 4. Driving the migration using Claude Code
- Open sourced their Gradle plugin
Upgrade philosophy
- Better to upgrade sooner, but that means need to do it more often
- Want to make it as painless as possible; tooling helps
DGS – GraphQL
- DGS framework – open sourced – https://netflix.github.io/dgs for making it easy to use GraphQL
- scheme to query data
- gRPC – performance. think “methods” not “data”
- Use graphql or gRPC over REST
Garbage Collection – ZGC
- Now default to ZGC
- Thought ZGC would be better for some use cases put turned out to be for all workloads
- Spend more CPU to lower pause times
- Mostly Java 21 and 25
VIrtual Threads
- Rolledback because wasn’t safe for all of their use cases.
- Java 25 fixed those problems so recently started looking at again.
My take
I’ve seen prior version of this talk. I was super curious to see how AI changed it. Unfortunately, that topic was last and I had to leave early to setup for my presentation (in a different building). It was nice hearing the plans to use Claude Code for the Spring Boot 4 upgrade. THere were other changes I did get to hear about so worth going!