M3

What M3 is

M3 is an open-source metrics engine for scaling Prometheus-based monitoring and centralizing metrics storage. Its M3DB component is a distributed time-series database for real-time metrics with long retention. The project supports PromQL, Graphite, and M3QL query languages, and accepts Prometheus remote write, StatsD, Carbon plaintext, InfluxDB line protocol, and JSON ingestion. Documented integrations include Prometheus, Graphite, Grafana, and InfluxDB. M3 supports horizontal scaling and retention policies. Its home page says it has been used to store tens of billions of active metric time series; its reliability design uses three data replicas with quorum writes and reads. The project reports an 11X compression ratio. M3 can be installed using binaries or Docker, and its documentation covers cluster deployment with Kubernetes. It supports Linux, macOS, self-hosted, and API platforms. M3 is open-sourced under Apache 2, and contributions are welcomed. The project began as an open-source metrics platform developed by Uber’s Observability team in 2014; community support is linked through Slack and GitHub. M3DB is not intended as a general-purpose time-series database, large-scale backfilling is not currently possible, and stored values are optimized for float64.

Who it suits

M3 suits teams scaling Prometheus-based monitoring and centralizing long-retention metrics, especially those using PromQL, Graphite, or its supported ingestion formats. Its horizontal scaling, deployment options, and documented integrations fit self-hosted metrics infrastructure.

Limits

M3DB is not designed as a general-purpose time-series database. Large-scale backfilling is not currently possible, and stored values are optimized for float64.

Getting started

1. Visit the M3 website.
2. Choose a deployment route using binaries or Docker.
3. For a cluster deployment, follow the Kubernetes guidance in the documentation.
4. Connect supported metrics sources and query tools as needed.

Verdict

Pick M3 for open-source, Prometheus-compatible metrics storage when long retention, scaling, and its supported query and ingestion options fit your needs. Look elsewhere for general-purpose time-series storage or if your workflow depends on large-scale backfilling or stored values beyond the float64 optimization.