Tail latency
The latency of the slowest requests (e.g., p99 or p95) — much more important to user experience than average latency.
Think of it like
One slow checkout line ruins the shopping trip even if 99 others are fast.
Example
Average latency is 50ms, but p99 is 500ms — some users wait 10x longer, which feels broken.
How it actually works
Average latency can be misleading. If 99% of requests are 50ms and 1% are 500ms, the average looks good but users in that tail suffer. Tail latency is caused by GC pauses, OS scheduling, cache misses, or congestion. Reducing tail latency often means: better hardware isolation, fast autoscaling, or shedding load.
For product teams
P99 latency determines user experience — focus on it, not average.
For engineers
Profile and eliminate sources of variance; prioritize speed over average efficiency.
Related
- P99 latency — The 99th percentile metric.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome