OTel metrics pricing and storage
OpenTelemetry metrics in CloudWatch use a per-GB ingestion model with included storage and query
access. There are no per-metric, per-API-call, or separate storage charges. For current pricing
details, see Amazon CloudWatch
Pricing
Key pricing characteristics
The OTel metrics pricing model has the following characteristics:
-
Per-GB ingestion — You pay based on the volume of data ingested, not the number of unique metric series.
-
Storage included — 15 months retention at no additional charge.
-
Console queries free — PromQL queries in the CloudWatch console (Query Studio, dashboards) are free.
-
API queries charged per sample — Programmatic PromQL queries are charged per million samples scanned.
-
No per-metric charge — Unlike Classic pricing, there is no cost per unique metric name or label combination.
How OTel compares to Classic pricing
The following table compares OTel metrics pricing with Classic metrics pricing.
| Feature | OTel Metrics | Classic Metrics |
|---|---|---|
Billing unit |
Per GB ingested |
Per metric per month |
Storage |
15 months included |
15 months included |
Console queries |
Free |
Free |
API queries |
Per million samples scanned |
Per 1,000 metrics requested (GetMetricData) |
High cardinality cost |
Linear with data volume (not metric count) |
Expensive — each unique dimension combination is a separate billable metric |
What counts as ingested data
Each OTLP data point contributes its serialized size to the ingestion volume. The serialized size includes the following components:
-
Metric name
-
All labels (keys and values)
-
Resource attributes
-
Timestamp and value
Larger label sets produce more bytes per data point. You can monitor your ingestion volume in the CloudWatch console under Usage Metrics.
Cost optimization tips
Use the following strategies to optimize your OTel metrics costs:
-
Use shorter label values where possible. Label values contribute to the byte size of each data point.
-
Drop high-cardinality labels that you don't query (for example, request IDs) at the collector level.
-
For Container Insights OTel, the default collection generates a predictable volume based on cluster size.
