Unexpected monitoring results

A reader's guide to results that look wrong on a provider's page, and usually are not.

Several things on a provider's monitoring look like faults and are not. This page is the list.

"Not enough data" on a provider with a long history

Not a contradiction. The history is what was measured; the verdict is about now. If too few independent workers are currently reporting on that provider, there is no confident current verdict, and the page says so rather than presenting an old one as current.

A ninety-day uptime figure that is missing

The provider has not been monitored for ninety days. The tile is shown, so you know the metric exists, and says how much of the window exists instead of averaging twelve days under a ninety-day label.

Grey cells in the availability timeline

Days nobody measured: before monitoring began, or a gap in reporting. Distinct from the colour used for a day the network measured and could not reach a verdict on. The legend names both, and neither is an outage.

A region missing from the breakdown

Nobody is running a worker there. A missing region is a blind spot in the measurement, not a finding about the provider.

Excellent uptime and poor reliability reviews

These measure different failures. Monitoring sees whether the network answers; it does not see storage failures, hypervisor problems, or badly handled maintenance, all of which happen with the network perfectly reachable. See Reliability.

A latency figure that seems high for a provider near you

The figure is measured from the workers, not from you. It is the median across wherever the network happens to have vantage points, so it is most useful compared against the same provider over time or against providers measured from a similar spread.

Two providers with the same brand and different figures

Probably different autonomous systems: a reseller and its upstream, or a group with separate networks. Every figure names the network it describes for exactly this reason.

A provider with no monitoring at all

It has not opted in. That is a choice about measurement rather than a judgement about reliability, and no figure is estimated to fill the space.

Simulated data warnings

Only ever seen on a development environment that is not connected to the probe network. If you see one in production, tell us: it means something is misconfigured, and it is badged rather than hidden so that it cannot be mistaken for measurement.