Crate p99 

Low-cost performance percentile histograms for Rust.

p99 provides Histogram, a fixed-size histogram with 64 logarithmic power-of-two buckets. It records event durations and approximates percentile values without retaining every observation.

Installation

Reference the crate in Cargo.toml in the usual way:

p99 = { version = "0.0.4" }

Enable the optional binary-scaling feature to use fixed-point arithmetic when calculating the target rank for integer percentiles:

p99 = { version = "0.0.4", features = ["binary-scaling"] }

Components

Histogram

A histogram accepts durations as std::time::Duration values or as integer counts in nanoseconds, microseconds, milliseconds, or seconds. It supports clearing the measurements, reading counts and minimum, maximum, and total duration values, and checking for arithmetic overflow.

Percentiles are approximated by interpolating within the matching logarithmic bucket. The generic Histogram::value_at_percentile method accepts a percentage from 0.0 through 100.0; convenience methods provide p50, p75, p90, p95, p99, p99.5, p99.9, p99.99, p99.999, and p99.9999.

Examples

use p99::Histogram;
use std::time::Duration;

let mut histogram = Histogram::default();
histogram.push_event_duration(Duration::from_micros(5));
histogram.push_event_time_ns(150);

assert_eq!(2, histogram.event_count());
assert_eq!(Some(150), histogram.min_event_time());
assert!(histogram.value_at_p99().is_some());

See the project README for further information.

Structs

Histogram

Low-cost performance percentile histogram using 64-buckets.