Introducing native histogram support in Amazon Managed Service for Prometheus

This article introduces native histogram support in Amazon Managed Service for Prometheus, enabling more efficient distribution tracking for Kubernetes and microservices workloads.

• Native histograms store entire distributions in single time series using exponential bucketing
• Reduces cardinality by up to 75% compared to classic histograms with fixed bucket boundaries
• Provides higher-resolution percentile calculations without predefined bucket boundaries
• Each populated bucket metered at 0.25 samples versus 22 series for classic 20-bucket histograms
• Fully compatible with existing PromQL queries and incremental adoption alongside classic histograms
• CloudWatch vended metrics track native histogram ingestion, active series, and cost drivers
• No new pricing dimension; standard Amazon Managed Service for Prometheus rates apply

Native histograms enable DevOps teams to achieve significant cost savings and improved tail-latency accuracy while maintaining full Prometheus compatibility.