>_

Latency Simulator

Analyze p50, p95, and p99 latency percentiles and tail-latency distributions.

p50

50ms

p95

185ms

p99

297ms

Tail ×

5.9×

Latency distribution

min

3.2ms

mean

68ms

p50

50ms

p90

138ms

p95

185ms

p99

297ms

How this simulator works

A latency distribution is not the average — it is the tail. Draw samples from a log-normal model and read p50/p95/p99 directly off the histogram.

Why percentiles, not averages

The mean is dominated by the long tail and hides the experience of most users. p50 is the median request; p99 is what your slowest 1% of requests feel.

Tail latency

  • Tail ratio = p99 / p50. A ratio of 10 means the worst 1% are 10× the median.
  • A heavier tail (higher spread) inflates p99 far more than p50.
  • Capacity planning that ignores p99 ships outages under load.