This presentation by Stefano Doni (Akamas CEO) at Performance Summit 2021 in London illustrates how AI-based optimization provides a more effective approach to Java Tuning that also helps to debunk some common JVM tuning myths based on established industry approaches and beliefs.

Key questions and debunked myths:

  1. Garbage Collector tuning: do the industry guidelines and metrics always lead to better application performance?
  2. Tradeoff among Latency, Throughput, Footprint: is there a way to improve these aspects at the same time?
  3. JVM tuning or not: is it worth tuning the JVM, or is it good enough for ensuring better performance to simply use the latest and greatest JVM?
  4. JVM tuning once for all: is it really a good practice to copy & paste known “good” VM configurations across different applications?