Cortex review
A self-hosted, open-source choice for multi-tenant Prometheus metrics at scale.
Reviewed by iTechGuides Editors · Editorial team · Updated Oct 2026
Cortex is an open-source, self-hosted time-series metrics storage system for organizations running Prometheus or OpenTelemetry at scale. It is designed for multi-tenant environments that need horizontally scalable ingestion and querying, with data and query isolation between tenants. Prometheus remote write and OpenTelemetry Protocol metrics ingestion bring data into the system, while PromQL provides access to stored metrics. Cortex supports Linux, macOS, HTTP APIs, and self-hosted deployments, with single-binary and microservices modes for different operating contexts.
Its strongest fit is a metrics architecture that needs long-term retention beyond local Prometheus storage. Cortex stores blocks in Amazon S3, Google Cloud Storage, Microsoft Azure Storage, or OpenStack Swift, and supports configurable per-tenant retention. High-availability replication across machines adds continuity for distributed deployments, while the horizontally scalable cluster architecture supports growth in ingestion and querying. Prometheus and the OpenTelemetry Collector are the central ecosystem connections, making Cortex more focused on observability metrics than on general-purpose time-series workloads.
Cortex is most appropriate for teams that want to operate their own multi-tenant Prometheus storage layer and are comfortable managing deployment, object storage, replication, and retention settings. The single-binary mode supports local testing and development, while the microservices mode provides a more distributed deployment shape. Its documented query model is PromQL, so teams seeking SQL access or a vendor-hosted database should choose another product. Choose Cortex when Prometheus-compatible metrics storage, tenant isolation, and scalable self-hosting are the priorities; look elsewhere when managed operations or broader query-language support matters more.
Cortex pros and cons
- Where it wins
- Multi-tenant isolation separates data and queries across tenants
- PromQL, remote write, and OTLP support cover core metrics workflows
- Object storage, replication, and retention support long-term storage
- Where it doesn't
- Self-hosted deployment leaves operations and infrastructure to your team
- PromQL is the documented query language rather than SQL
- Microservices deployment can demand more coordination than a single binary
Cortex fact sheet, pricing and score →
Advertiser disclosure: iTechGuides is reader-supported. We may earn a commission when you click some links. How we rank.
Last updated · How we research and update