Capacity Curve

Find the point where more RPS causes latency to explode.

capacity

Detailed Description

A capacity curve shows the relationship between concurrency, throughput, and latency. At first, RPS rises as concurrency increases; after saturation, latency climbs sharply and errors may appear.

The safe operating point is before the knee of the curve. Use that point to set autoscaling thresholds, queue limits, worker concurrency, and per-instance traffic budgets.

Mental Model

increase concurrency
-> plot RPS + p99
-> find knee of curve

When To Use

->Decide instance count before launch
->Choose autoscaling thresholds
->Find max safe concurrency per pod

Tools: autocannon matrix, Grafana dashboard, SLO capacity planning

capacity-loop.sh

Live Editor

Output

Run code to see output...