Performance Issues

Measure the bottleneck in the actual workload

Record the operation, input size, concurrency, application version, and latency before changing code. Separate database time, external API time, application CPU, memory pressure, and rendering. Inspect query plans and indexes for the real filter and ordering.

Use bounded data sizes and the host JVM/database profiling tools. A generic framework benchmark is not a capacity guarantee. Repeat the same workload after a change and compare latency, throughput, errors, and resource use. See capacity planning.