Measurements

A measurement is a scheduled test — ping, traceroute, DNS, SSL/TLS, HTTP or NTP — run from a set of probes toward a target. It has an owner, a type, a target, an interval, a start and optional stop time, and a probe selection.

Status, and why it decides everything

A measurement is either ongoing or in a terminal status — stopped, forced to stop, failed, archived, or unable to find suitable probes. That distinction drives how Atlas treats it:

• Ongoing measurements are always read live from RIPE. Their results are still changing, so caching them would only serve you something stale.
• Terminal measurements are cached locally, one record per (measurement, probe) pair. Their results are never going to change, so they are worth keeping — and they load far faster than a round trip upstream.

If you are wondering why one measurement's detail page is instant and another takes a moment, this is why.

Measurement metadata is cached read-through with a short lifetime for everything, ongoing or not: a local copy is served while it is fresh, and refreshed from RIPE when it isn't.

Types

Type	What it tells you
ping	Round-trip time and packet loss to a target
traceroute	The path, hop by hop
DNS	What a resolver or authoritative server answers
SSL/TLS	Certificate and handshake details as seen from each probe
HTTP	Status and timing of an HTTP request
NTP	Clock offset and delay

For DNS measurements Atlas additionally decodes the raw wire-format answer into a structured form, while keeping the original verbatim — so the decoding can be redone later without going back to RIPE.

Built-in, well-known and user-defined

Built-in measurements are run by RIPE itself and readable by everyone. Two of them are load-bearing for Atlas: measurement 7000 (connection events) and 7001 (uptime).

Well-known measurements are the ones RIPE tags as well-known targets — the standing measurements against significant infrastructure that the rest of the platform leans on. Atlas refreshes this set hourly and gives it its own page, because "what is being measured that everyone can read" is a common starting question.

User-defined measurements (UDMs) are created by individual users and cost that user credits. Some are public, some private; a private measurement's metadata is not readable, and Atlas records that fact rather than retrying.

Participating probes

The probe set is chosen when the measurement is created and drifts afterwards as probes disconnect and are replaced. So "which probes are in this measurement" is a question with a current answer, not a fixed one — and a measurement whose participant count has quietly fallen is a common cause of thinning results.

The anchoring mesh is the extreme case of this: see Mesh health, which exists precisely to catch participation gaps.

See the Measurements guide for the pages built on all this.