Glossary

p99 latency

p99 latency is the 99th percentile of request durations: 99% of requests complete faster than this value, and 1 in 100 is slower. It is the standard metric for latency SLOs in production systems because it captures the experience of users who encounter the system under stress — during GC pauses, cache misses, lock contention, or network retransmissions — without being distorted by the extreme outliers that p99.9 captures.

Formula: With N sequential calls each at p99 latency l_i, combined p99 ≈ sum(l_i); probability of hitting at least one slow call grows with N.

Last updated: March 2026