JProfiler Get rid of your performance problems and memory leaks! Modern Garbage Collection Tuning Java Performance Training Courses COURSES AVAILABLE NOW. We can provide training courses to handle all your Java performance needs Java Performance Tuning, 2nd ed The classic and most comprehensive book on tuning Java Java Performance Tuning Newsletter Your source of Java performance news. Subscribe now! Enter email: Training online Threading Essentials course JProfiler Get rid of your performance problems and memory leaks!

Back to newsletter 296 contents

The most interesting item this month for me was from Netflix's talk (see news items below) about their use of Java in 2025, and specifically how that relates to their experience with reactive programming and their intention to replace that with virtual threads.

This is fascinating. Netflix was one of the originators of reactive frameworks in Java, and they're saying that reactive leads to too complex code, too hard to debug and too hard to maintain. They have tried out virtual threads as a replacement for handling concurrency, and found them effective but had to hold off when they found pinning was a real problem. But pinning is fixed in Java 24 and now Netflix are going to push ahead with virtual threads.

Now on to all the usual newsletter list of links, tips, tools, news and talks, and as usual I've extracted all the tips into this month's tips page.

A note from this newsletter's sponsor

News

Java performance tuning related news

• How Netflix Uses Java - 2025 Edition. Relevant for performance: upgrading JVM from 8 to 17 gave a "free" 20% decrease in CPU used by GC; ZGC in JDK 21 moved their highest volume services from 1 second GC pauses on G1 to no significant GC pauses; they found RX Java added too much complexity to code and debugging so are backing out of reactive programming; they expect virtual threads to completely replace reactive; virtual threads had deadlocks from pinning synchronized blocks until JDK 24, but from JDK 24 virtual threads are fine; they experimented with native image for faster startup but find it "too hard to get it right" and it worsens the development experience; REST provides too much data, graphQL and gRPC are preferred
• -Xlog:class+load+cause", "-XX:LogClassLoadingCauseFor=SOME_CLASS - print a stack trace on class loading
• Interesting but not directly useful description of how JFR will take safepoint traces but (possibly) avoid safepoint bias

Tools

Java performance tuning related tools

• Collecting and transporting a JFR dump from a containerized environment (uses openshift oc command but translates directly to Kubernetes kubectl command)
• Benchmarking and profiling Java with JMH
• GCSee uses GCToolKit to create simplified views of various GCs

Articles

• GC Tuning: A Masterpiece in Performance Engineering (Ram Lakshmanan/Bulgarian Java User Group)
• Continuous Profiling: The Missing Piece in Your Observability Puzzle (Mohammed Aboullaite/Code Remix)
• Garbage Collection In Java (Sagar V Hande/Scripting with Ease)

Jack Shirazi

Back to newsletter 296 contents