RIPE Atlas in five minutes

Everything in Atlas is built on RIPE Atlas, so a handful of its terms appear on almost every page. This is the short version; RIPE's own documentation goes deeper on all of it.

Probes

A probe is a small device a volunteer runs on their own network. It has a stable numeric ID (probe 6042), an IPv4 and/or IPv6 address, an ASN and prefix per address family, a location, a firmware version, and a set of tags.

Probes are volunteer hardware on volunteer networks, which is the single most important thing to remember about the data: they get unplugged, re-homed, renumbered and forgotten. A probe's status moves through never connected → connected → disconnected, and eventually to abandoned when it has been down long enough that nobody expects it back.

Atlas archives all of that, so a probe here is its whole history rather than just its current row. See Probes.

Anchors

An anchor is a bigger, better-connected probe, typically hosted in a data centre rather than someone's flat. Anchors do everything probes do, and they additionally serve as targets — well-known endpoints that other probes and anchors measure toward.

That second role is what makes them structurally important: the anchoring mesh, DNSMON, and most comparative latency work all point at anchors. See Anchors and The anchoring mesh.

Measurements

A measurement is a scheduled test run from a set of probes toward a target. The types are ping, traceroute, DNS, SSL/TLS, HTTP and NTP. A measurement has an owner, an interval, a start and optional stop time, and a probe selection that determines who runs it.

Two distinctions matter when reading Atlas:

• Built-in versus user-defined. RIPE runs a standing set of built-in measurements — the anchoring mesh, DNSMON, and similar — that everyone can read. A user-defined measurement (UDM) is one somebody created themselves.
• One-off versus recurring. A one-off runs once; a recurring measurement keeps producing results on its interval until it is stopped.

Running measurements costs credits, which you earn by hosting probes and spend by measuring. Only your dashboard deals in credits — the rest of Atlas reads results without spending anything. See Measurements.

Connection events

Probes report to RIPE's infrastructure continuously. Each connect and disconnect is a connection event, and the stream of them is how uptime, outages and coverage are reconstructed after the fact. Most of the outage and trend analysis in Atlas eventually reduces to counting these. See Connection events and uptime.

Where Atlas differs

Atlas does not redefine any of these terms — it reuses RIPE's vocabulary directly. What it adds is history, cross-referencing against other data sources, and fleet-level aggregation. The one place to be careful is API keys: a RIPE Atlas API key and an Atlas API key are unrelated things. See Accounts and roles.