Lexicon
Metric
A metric is a named numeric measurement tracked over time, with a type, unit, and attributes. How metric types shape aggregation, and where costs hide.
Definition
A metric is a named numeric measurement tracked over time, request counts, memory usage, response latency, carrying a type, a unit, and attributes that slice it into series. Metrics are the cheapest signal to store and the fastest to query, which makes them the workhorse of dashboards and alerting.
What it means in observability
Metrics trade detail for efficiency. Where a log records every event and a trace records every request, a metric records aggregates, and that compression is what lets years of history sit behind a chart that renders instantly. The craft lies in reading them correctly, because a metric's type determines which questions it can answer: the same aggregation that summarizes one type correctly renders another meaningless, which is why type confusion produces confident dashboards of nonsense.
How it works in practice
Three types cover most of practice. Counters only accumulate, counting events like requests or errors, and are read through rates and increases. Gauges snapshot a current value that moves both directions, memory in use, queue depth. Histograms capture distributions, powering the latency percentiles everyone alerts on. Around the types sit temporality, whether points report deltas or running totals, and attributes, which multiply one metric name into many series and drive cardinality.
Where it gets hard
Cardinality is the cost center: one unbounded attribute turns a tidy metric into millions of series, and per series pricing turns that into invoice drama. Aggregation mistakes are the correctness center, averaging percentiles, summing gauges, plotting raw counter values, each a classic. And metrics alone explain nothing; they tell you something changed, and the investigation still needs the traces and logs behind the number.
Where Tsuga fits
Tsuga's metric explorer surfaces each metric's type, temporality, and supported aggregations so queries start correct, and flat per GB pricing takes the cardinality panic out of keeping the attributes you need.
Related terms
- AWS CloudWatchAmazon CloudWatch is AWS's built in monitoring service: metrics, logs, alarms, and dashboards, with tracing through X-Ray, integrated by default with nearly every AWS service.
- CardinalityCardinality is the number of distinct values, or distinct value combinations, that an attribute or set of attributes can take.
- CounterA counter is a metric type for values that only accumulate: requests served, errors thrown, bytes transmitted, jobs completed.
- GaugeA gauge is a metric type that records a point in time value that can move in both directions: memory in use, queue depth, active connections, temperature, items in a cart.
- GrafanaGrafana is the open source standard for dashboards, and Grafana Labs builds an observability stack around it: Loki for logs, Tempo for traces, Mimir for metrics, with Grafana Cloud as the managed offering.
- HistogramA histogram is a metric type that captures how values distribute, by counting observations into buckets, rather than recording what any single value was.
- Infrastructure monitoringInfrastructure monitoring is the practice of tracking the health, performance, and capacity of the layer applications run on: hosts, virtual machines, containers, Kubernetes clusters, and managed cloud services.
- Semantic conventionsSemantic conventions are OpenTelemetry's standard vocabulary: the agreed names, types, and values for the attributes that describe telemetry.
- SLIAn SLI, service level indicator, is the measurement behind an SLO: the specific, computable signal that stands in for whether users are having a good experience.
- TelemetryTelemetry is the data software and infrastructure emit about their own behavior.