observability
Histogram
Track latency distributions with bucket counts. Compute percentiles for SLA monitoring.
count0
avg0ms
p500ms
p950ms
p990ms
≤10
≤25
≤50
≤100
≤250
≤500
≤1000
≤2500
≤5000
>5000
latency (ms)
// recent observations (last 20)
none yet
How It Works
- Each observation falls into a bucket (≤ threshold)
- Percentiles estimated from cumulative distribution
- p50 = median, p99 = tail latency
- O(1) per observation, O(buckets) for percentile
Use Cases
- Request latency tracking (Prometheus)
- SLA monitoring (p99 < 200ms)
- Database query performance
- Response size distributions