The Limit of Fixed Rates
A fixed sample rate is a poor compromise. Set it high and a traffic spike floods your backend. Set it low and rare endpoints almost never get sampled. Adaptive sampling adjusts the rate automatically.
How It Adapts
The idea is to target a budget, such as a number of traces per second, rather than a fixed percentage:
- Per operation rates: a busy endpoint gets a low rate, a rare endpoint a high one, so both stay represented.
- Feedback loops: the sampler watches recent throughput and nudges rates to hit the budget.
- Guaranteed minimums: low traffic paths get a floor so they are never invisible.
Why It Helps
The result is steady backend cost with fair coverage. Hot paths do not drown out the rare ones, and a sudden surge does not blow the budget.
Key idea
Adaptive sampling tunes the rate per operation toward a trace budget, keeping cost stable and rare endpoints visible under changing load.