Cardinality
The number of unique label combinations in a metric — high cardinality is a common scaling challenge.
Definition
Cardinality refers to the number of unique time series produced by a metric. Each unique combination of label values creates a separate time series, so adding labels with many possible values — such as user IDs or request IDs — can cause an exponential explosion of series.
In depth
High cardinality is one of the most common performance issues in Prometheus-based monitoring. A single metric with four labels, each with 100 possible values, can produce up to 100⁴ = 100 million time series. Most TSDB engines degrade significantly at these scales. Strategies to manage cardinality include aggregating high-cardinality labels before ingestion, using exemplars to sample individual values while still tracking aggregate metrics, and applying recording rules to pre-compute expensive aggregations. obseria.io's adaptive cardinality management automatically identifies and caps runaway series before they impact query performance.
Related terms
See Cardinality in action
obseria.io gives you full-stack observability — logs, metrics, traces, and AI-powered root cause analysis.