Statistics

Statistics answer "how many, right now". Trends answer "how has that moved". Same underlying data, different question — if you find yourself reading a date range off a Statistics page, you probably want Trends.

Anchor stats

/statistics/anchors · public

Aggregations over the anchor fleet, using the same filter vocabulary as the anchor list. The aggregation is computed in the browser over whatever the filters returned — so what you see is a summary of the filtered subset, not of the whole fleet with a filter applied afterwards. Widen the filters to widen the statistics.

Probe stats

/statistics/probes · public

The same idea for the whole probe fleet: aggregate counts over public probe metadata — status, country, ASN, firmware and the rest of the standard breakdowns.

Note that RIPE's probe status IDs are a different vocabulary from measurement status IDs. They share the word "status" and nothing else.

Connection-event stats

/statistics/connection-events · public

A window-scoped dashboard over the daily connection-event rollups. The window can be:

• a rolling preset — the last 7, 30 or 90 days
• a single UTC day
• a free range — any start and end

Rolling windows are "today minus N", which means today itself is partial. For a clean comparison between periods, use explicit days or a range.

Data delay

/statistics/data-delay · public

RIPE's own ingestion delay — how long results coming off probes take to land upstream. See Data delay and freshness for what the metric actually measures and why it matters to everything else.

Three window modes, and the difference between them is not cosmetic:

• Quick — a rolling 1h, 6h, 24h or 7d window, with auto-refresh on at a 60-second tick. This is the "is it healthy right now" mode.
• Day — one specific UTC calendar day, auto-refresh off, because historical data isn't moving and re-polling it would only cost you.
• Range — a custom UTC start and end.

The chart supports drag and scroll zoom on the time axis, which is what you want when reconstructing a specific bad hour.

Header tiles show the current delay alongside the percentile spread. Watch the gap between median and p99 rather than the headline: a stable median with a climbing p99 means a subset of probes is struggling, not the whole pipeline.

Missed disconnects

/statistics/missed-disconnects · staff

A diagnostic for a specific data-quality problem: RIPE sometimes drops a disconnect event from measurement 7000. The tell is two consecutive connect events for one probe with no disconnect between them — the first session's disconnect was lost.

This matters because naive uptime treats the whole gap as connected time, so missed disconnects silently inflate availability figures.

The report breaks them down by controller, country, ASN, probe type and controller handoff, and — importantly — reports a rate (missed ÷ (missed + disconnects)) alongside the raw count. A big controller will naturally have the most missed events; the rate is what tells you whether it is actually worse than the others.

A toggle switches between the full report and a flapper-excluded slice, which drops probes that connect and disconnect constantly and would otherwise dominate the numbers. The switch is instant — both slices come back in one scan.

---

Background reading: Data delay · Connection events and uptime