Uptime
Probe
Probe is a single check run against a service from one place at one moment. Monitoring is a stream of probes.
How it is measured
Each probe returns a result (pass or fail), a timing breakdown, and metadata: region, protocol, and target. Frequency, locations, and timeout belong with the number.
Combine probes with a rule, like failing in two of three regions on two consecutive runs. One probe is an observation. A rule turns observations into an alert.
Worked example
A monitor probes a travel blog every 60 seconds from Tokyo, Dallas, and Frankfurt. At 12:00:30 Tokyo reports a connect timeout, while Dallas and Frankfurt return 200 in 310 and 280 ms. At 12:01:30 Tokyo succeeds.
That is one failed probe out of 360 that hour. Without a rule it would have paged. With a two-region rule it is a dot on a graph, probably a transit hiccup near Tokyo.
How it differs
Probe is one check from one place. HTTP check is a type of probe that makes a web request. The probe is the unit. HTTP check is one kind, and TCP or DNS probes work differently.
Common errors
Probing from the same data center as the server. Using one location. Using too long an interval. Setting 30-second timeouts that hide slowness. Treating a single failure as an outage. Using a probe that is cache-served.
In practice
Check where your probes originate and whether they hit the CDN or the origin. Use at least three regions. Write the alerting rule in terms of probe results, not feelings.