Latency Budget

API latency is usually the sum of network, proxy, application, serialization, database, cache, and downstream dependency time.

Do not optimize blindly. Measure where time is spent.

A request taking 500 ms may consist of 20 ms application logic, 300 ms database queries, 100 ms external API calls, and 80 ms network/proxy overhead.

Profiling and tracing reveal the actual bottleneck.