How Can Machine Learning Assist to Improve JVM Performance

This conference talk explores how machine learning can be leveraged to automatically tune JVM flags for improved Java application performance. Discover an approach for auto-tuning JVM flags studied in collaboration with Oracle, where five different machine learning models were trained on data extracted from G1 GC logs to enhance throughput and latency. Learn why many developers avoid using JVM flags despite their potential to significantly boost performance, especially in memory-related configurations. The presentation offers promising results and practical tools that can make the optimization process both beneficial and enjoyable for developers struggling with JVM performance tuning. The research is based on the speaker's master's thesis conducted with Oracle's Java Platform Group and was presented at Jfokus 2025.