P50 / P95 / P99 Latency
Also known as: percentile latency, tail latency
Percentile measurements of response time. P95 = 95% of requests complete in this time or faster. P99 captures the slow tail that average latency hides — often where the worst user experiences live.
Source
- Google SRE Book — Monitoring Distributed Systems
Related terms
Uptime
The percentage of time a service is reachable and responding correctly over a given window.
Downtime
Any period during which a service fails to respond, returns errors, or is so degraded it is unusable.
SLA (Service Level Agreement)
A formal contract between a service provider and a customer that defines minimum service levels and the con…
SLO (Service Level Objective)
An internal reliability target — for example, "99.9% of API requests succeed within 500 ms over a rolling 3…
SLI (Service Level Indicator)
The raw measurement that an SLO is calculated from — request success ratio, latency percentile, or freshnes…
MTTR (Mean Time To Resolve / Recover)
Average time between an incident being detected and the service being fully restored.