Uploaded April 2026 | Updated September 2026, 3 weeks ago
Please subscribe to our YouTube channel @ youtube.com/@DevoxxForever
Subscribe to LinkedIn @ linkedin.com/company/voxxed-days-amsterdam
Follow us on Twitter @ twitter.com/voxxedamsterdam
Upgrading to a new Java version is often driven by the desire for new language features, or the need to stay current with security updates. In this talk, we show how moving to Java 25 became a turning point for fixing severe tail-latency and GC instability in a high-throughput production system at Picnic.
This session presents a real production case study where upgrading to Java 25 triggered a deeper rethink of JVM memory usage and garbage collection. Earlier trial with ZGC resulted in startup failures and worse tail latencies due to higher memory usage. Java 25’s Compact Object Headers enabled a counter-intuitive shift: shrinking the heap to re-enable Compressed OOPs and improve object density. Combined with targeted memory optimizations in code, it led to more predictable GC behavior and significantly stable latency.
Key takeaways
Practical mental models for JVM memory behavior
When large heaps hurt latency more than they help
Concrete techniques to reduce tail latency without adding hardware
Target audience
Intermediate to advanced Java developers
Backend engineers working on high-throughput systems
Platform engineers responsible for performance and reliability
Please subscribe to our YouTube channel @ youtube.com/@DevoxxForever
Subscribe to LinkedIn @ linkedin.com/company/voxxed-days-amsterdam
Follow us on Twitter @ twitter.com/voxxedamsterdam
Upgrading to a new Java version is often driven by the desire for new language features, or the need to stay current with security updates. In this talk, we show how moving to Java 25 became a turning point for fixing severe tail-latency and GC instability in a high-throughput production system at Picnic.
This session presents a real production case study where upgrading to Java 25 triggered a deeper rethink of JVM memory usage and garbage collection. Earlier trial with ZGC resulted in startup failures and worse tail latencies due to higher memory usage. Java 25’s Compact Object Headers enabled a counter-intuitive shift: shrinking the heap to re-enable Compressed OOPs and improve object density. Combined with targeted memory optimizations in code, it led to more predictable GC behavior and significantly stable latency.
Key takeaways
Practical mental models for JVM memory behavior
When large heaps hurt latency more than they help
Concrete techniques to reduce tail latency without adding hardware
Target audience
Intermediate to advanced Java developers
Backend engineers working on high-throughput systems
Platform engineers responsible for performance and reliability
