Java Performance in 2021: How AI Optimization Will Debunk 4 Long-standing Java Tuning Myths

The main rules and best practices we daily use (and our tools recommend) to analyze and optimize the Java performance are no longer working. New approaches – like AI – are needed to gain back the performance wins JVMs offer

Java is ubiquitous in online services, yet ensuring Java applications’ availability and performance remains a challenging task. In this talk, we show how established industry approaches and widely-accepted beliefs about Java tuning are wrong and how AI breaks through long-standing limitations. We will be looking at the surprising results AI achieved when tuning 100s of Java apps and we will debunk four common JVM tuning myths:

1) Garbage collector tuning: the industry uses specific guidelines and metrics to tune the GC, but are you really sure they lead to better application performance?

2) The Latency – Throughput – Footprint trade-off: is it a law of physics, or there are ways to achieve all of them at the same time?

3) Let the VM do it: why don’t you simply use the latest and greatest JVM and simply forget about JVM tuning?

4) Cargo cult JVM tuning: developers often copy and paste “good” JVM configurations across different applications, is this strategy actually effective?



I’m obsessed about performance optimization and I lead the Akamas vision for autonomous performance optimization powered by AI. Prior to my role at Akamas, I led the Research & Development business unit at Moviri, focusing on creating innovative IT performance optimization products. Before that, I delivered dozens of successful capacity and performance management consulting projects at Moviri for major national and international enterprises. In 2015, I won the Best Paper award at the Computer Measurement Group international conference for my contributions to capacity planning and performance optimization of Java applications ( My interest: extracting top performance and efficiency from the IT stack. Tweets at: