A time series is an ordered sequence of metric values indexed by successive timestamps. In Prometheus, the metric name plus optional key–value labels uniquely identifies each series, making labels the dimensions used to slice measurements. section 4.2.1 section 5.3
A time-series database indexes observations primarily by time, usually in equal buckets, and answers questions about values or aggregates matching conditions across an interval. A single observation combines a timestamp, a series identity, and a value. section 5.3
API requests by dimensions
The series api_requests_total{method="GET", path="/some/interesting/thing"} represents API invocations for one HTTP method and path. A counter records the accumulated total; comparing values across periods gives the invocations during that period. Labels can be standardized during scraping when sources use different names. section 4.2.1 section 5.3
Time series support several observability tasks:
- Track system or application measurements, such as container memory usage or service invocations. section 5.3
- Connect aggregated metric values to related metrics, logs, or representative trace spans during investigation. section 11.1.3
