Skip to content

Deployments ​

The Deployments group answers one question from several angles: where is the network under-deployed? It is the group to reach for when you are making the case for a probe somewhere, or working out which gap to close first.

Most of it is staff-gated; the new-probes chart is public.

Country coverage ​

/deployments/country-coverage · staff

A live per-country rollup of the probe fleet — around 250 countries, returned in one call, with filtering and sorting done in the browser so it stays responsive.

Two lists carry the weight:

  • Gaps — countries with no connected probe at all. Dark.
  • Fragile — countries hanging on a single connected probe. One unplugged device from going dark.

The fragile list is the more actionable of the two: a country that has never had a probe is a long project, whereas a country one probe away from dark is something you can fix this month.

A condensed version of this — counts of dark and single-probe countries, plus recent transitions — appears on your dashboard, fed by an hourly detection pass.

AS population gaps ​

/deployments/as-gaps · staff

The same question at network rather than country granularity: which significant autonomous systems have no probes?

Two tabs over one filter set:

  • ASN gaps — the per-AS table, the default landing view
  • By country — a country rollup; clicking a row filters the ASN tab to it

The filters are what make it useful, and they are all reflected in the URL so a prioritised list is a shareable link:

FilterDefaultWhat it does
max_probes3Only ASes with at most this many probes
min_population10,000Only ASes with at least this many estimated users
country—Restrict to one country
statuses—Which probe statuses count as "having a probe"

The population figure is an estimate of users behind the AS, so the ranking is effectively users per probe — which is the number that makes the case for a deployment. Cross-check any individual candidate on its AS dossier.

The statuses filter deserves attention: an AS whose only probes are disconnected is arguably a gap, and whether you count it changes the list considerably.

Weather and outages ​

/weather · staff

No longer collected

Weather collection, the correlation report and the weather alerts were stopped on 4 October 2026: in their time they found nothing worth acting on. The page still shows everything they recorded, as it was left — nothing on it updates.

Correlates hourly per-country temperature against a per-country hourly disconnect signal derived from connection events.

The method is deliberately conservative, because both series have a strong daily cycle and a naive correlation between them would be spurious:

  1. Anomalies, not raw values. Each series has its hour-of-day mean subtracted before anything is correlated, so the shared daily rhythm cancels out.
  2. Lagged cross-correlation. Heat-related failures lag the peak — afternoon heat, evening load, power strain, probes drop — so the reported figure is the lag with the strongest correlation, not the same-hour one.
  3. Extreme-heat lift. The mean disconnect rate in the hottest 10% of hours divided by the overall mean: "when it's hot here, outages run N× baseline."

The page also carries the last per-country temperatures collected and the weather alerts that were raised.

WARNING

Correlation is not causation, and probe outages are a biased sample — they tell you about places with probes, on networks whose hosts leave them plugged in. Read everything here as a signal worth investigating, never as a claim about what happened.

New probes ​

/graphs/probes/new · public

Deployment rate over time: how many probes came online, filterable and viewable three ways.

  • Chart — the default, and the one that carries a save-as-image control for slide decks
  • Map — where the new probes are, loaded separately when you open the tab
  • Details — the underlying rows

The tab is reflected in the URL, so a deep link lands where you meant. The filter panel starts open and collapses itself once the first result arrives, leaving the active filters visible as chips; after that it stays wherever you put it.


Background reading: Probes · Connection events and uptime

Atlas — built on RIPE Atlas data.