p99 latency
The latency value that 99% of requests complete under. A common tail-latency target for production systems.
Definition
p99 latency is the latency value that 99% of requests complete under.
It is a tail metric, which means it captures the slow end of the distribution.
Why it matters
Users experience the tail.
If your p99 is bad, your system feels random even if the average looks fine.