Painful days spent optimizing the wrong part of a flow. Repeated attempts to blindly fix a performance issue. Speculative micro-"optimizations" that clutter the code with no measurable benefit. Only to later trace the problem to the most unexpected cause. This talk explains how to correctly zoom in on a performance bottleneck using distributed tracing, metrics, and method profiling. We learn to read a flame graph, then we review a series of bottlenecks typical for backend systems, like connection pool starvation, time-consuming Aspects, network calls, CPU hotpoints, lock contention, and Virtual Thread pinning. Join this session and prepare for the performance issues that any successful system will eventually have.

Tracing Java's Hidden Performance Traps by Victor Rentea