CMG Impact 2021: Java performance in 2021. How AI optimization will debunk 4 long-standing Java tuning myths

Stefano Doni (CTO at Akamas) will speak at the CMG Impact 2021 online conference on January 27th.

Over the years, a certain kind of conventional wisdom about Java performance tuning has taken hold of the development and performance engineering community:

1. Java intrinsically forces developers to clear-cut tradeoffs between memory footprint, throughput, and latency.
2. Tuning guided by JVM metrics alone might not yield the results you expect.
3. Newer JVM releases are always better than old ones when it comes to performance.
4. It’s easy to tune the JVM. 

So what is a performance engineer to do? AI-based technologies can help. An AI can take in all the relevant stack parameters, and metrics from monitoring tools, and generate target configurations that optimize for any given goal. That goal can be set to directly solve for high-level application performance target e.g. maximize payments/sec while keeping cost SLAs under a threshold. The emphasis shifts then from low-level JVM tuning to end-to-end, continuous high-level optimization.

During our session, we will show how Autonomous Performance Optimization managed to achieve impressive results, such as reducing CPU usage by 40%, while also improving performance in Renaissance, a Java-based benchmark developed by Oracle Labs.