If you run Kubernetes or microservices workloads on AWS, your metrics likely carry dozens of labels: namespace, pod, container, node, deployment, replica set, and custom business dimensions. To get a complete picture of your environment, you may be splitting your metrics pipeline: Amazon CloudWatch for AWS metrics, and a separate Prometheus-compatible backend for high-cardinality (many unique label combinations) container and application metrics. Some teams go further. They use Prometheus CloudWatch exporters to pull AWS resource metrics into their Prometheus backend through GetMetricData API calls. This adds operational overhead and cost but lets them query everything in one place.
Amazon CloudWatch now natively ingests OpenTelemetry metrics and supports querying them with Prometheus Query Language (PromQL). This preview capability introduces a high-cardinality metrics store that supports up to 150 labels per metric, so you can send rich, label-dense metrics…
https://aws.amazon.com/blogs/mt/introducing-opentelemetry-promql-support-in-amazon-cloudwatch/
