Network
Measurement regions
Latency is a property of a network path, not of a provider alone, so where a measurement was taken is part of the measurement. Every observation records the region it came from, and results are never averaged across regions without saying so.
Collecting now
| Region | Agents online | Samples / 24h | |
|---|---|---|---|
| Barcelona | 0 / 0 | none |
Why the vantage point is part of the measurement
A provider that answers in 40ms from Frankfurt can take 220ms from São Paulo, and neither figure is wrong. A single global ranking answers “fastest from where the measurement was taken” while appearing to answer “fastest”. Every observation therefore records the region it came from, and every ranking here can be filtered to one.
The figures above cover the last 24 hours and count successful responses only for the percentiles. A request that timed out is not a latency sample — it is counted against availability instead, which is why a region can show a low median and a low success rate at the same time.
The success rate is across every provider and method measured from that region, including ones that are known not to implement a call or that rate-limit deliberately. It is a property of the field being measured rather than of the region: a low figure here usually means several endpoints there refuse a method, and the per-provider tables are where that is attributed.