Profiling

Profiling measures runtime behavior such as CPU time, allocations, memory retention, lock contention, I/O waits, and call frequency. It replaces performance guesses with evidence.

Practical workflow

Start with a representative workload and an end-to-end symptom. Use sampling or instrumentation to locate the dominant cost, change one meaningful factor, and measure again. Optimize bottlenecks rather than code that merely looks inefficient.