query performance

This hands-on guide shows how to analyze and optimize AI API latency and throughput from the Linux shell—no APM needed. Using curl, jq, ab, GNU parallel, sysstat, iperf3, and time, it builds a repeatable harness to capture per-request timings, quantify p50/p95/p99, probe concurrency, watch CPU/GPU/I/O/network, sanity-check transport, and apply an optimization checklist, then codify performance baselines in CI.