Metadata Plane · Table & Technical Catalogues

Apache Polaris

Open-source catalogue that implements the Apache Iceberg REST catalogue API.

Overview

Apache Polaris is a catalogue for Apache Iceberg tables, built on the Iceberg REST protocol. Its purpose is stated plainly by the project: a vendor-neutral implementation of that specification, so a lakehouse is not tied to one platform's catalogue.

It holds catalogues, namespaces that can be nested, tables, views, policies, principals and roles. Iceberg is the primary format, and a beta generic tables API registers non-Iceberg tables such as Delta, CSV or Lance so they can at least be catalogued.

Access control is role-based with a deliberate indirection: privileges are granted to catalogue roles, and catalogue roles are granted to principal roles, rather than to people directly. Polaris then vends temporary storage credentials at query time, so engines never hold cloud storage keys.

Federation is a notable strength. A Polaris server can front other catalogues rather than replacing them, including remote Iceberg REST catalogues such as AWS Glue or another Polaris, a Hive metastore over Thrift, or BigQuery Metastore.

Table maintenance is expressed as policy rather than performed: system policies define data compaction, metadata compaction, orphan file removal and snapshot expiry, attachable to a table, namespace or catalogue, and engines act on them.

It was written at Snowflake, open-sourced in 2024, entered the Apache Incubator that August and graduated to a top-level project in early 2026. Release 1.7.0 arrived in August 2026, with engines including Spark, Trino, Flink, Doris, StarRocks and Dremio.

Key features and capabilities

The same headings are used for every table & technical catalogues entry, so two tools can be read side by side.

What it catalogues
  • Catalogues holding namespaces, which can be nested, then tables and views
  • Tables map to Iceberg, and generic tables cover other formats
  • Identity and authorisation entities: principals, principal roles, catalogue roles, privileges and policies
  • Catalogues are internal, or external and federated
Table formats
  • Apache Iceberg is the primary format, read and write
  • Generic tables, in beta, register Delta, CSV, Lance and others as metadata
  • Generic tables carry no Iceberg-style commit protocol
Engines that can use it
  • Named engines: Spark, Trino, Flink, Doris, StarRocks and Dremio
  • Client guides published for Spark, Trino and Flink, plus a dedicated Spark client
  • Snowflake and Dremio named as consumers
Access control and auditing
  • Role-based access control with an indirection through catalogue roles
  • Privileges cover table, view, namespace, catalogue and policy operations
  • Temporary storage credentials vended at query time, so engines hold no cloud keys
  • External identity providers, and external policy decision points such as Open Policy Agent
Interoperability
  • Implements the Iceberg REST catalogue API
  • Federates to remote Iceberg REST catalogues, including AWS Glue or another Polaris
  • Federates to a Hive metastore over Thrift, and to BigQuery Metastore
  • Published catalogue and management API specifications
Table maintenance and operations
  • Maintenance expressed as policy: compaction, metadata compaction, orphan file removal, snapshot expiry
  • Policies attach to a table, namespace or catalogue; engines carry out the work
  • Persistence: PostgreSQL for production, MongoDB experimental, in-memory for testing only
  • Scalability limits are not published, though a benchmarking tool is
How it runs
  • Self-hosted only; source, binaries, a Helm chart and Docker images per release
  • Deployment guides for AWS, Azure and Google Cloud
  • Storage guides for S3 and MinIO, Azure Blob, Google Cloud Storage, Ozone and Ceph

Pricing

Open source

Free under the Apache 2.0 licence. The foundation offers no managed or paid edition.

Vendor pricing page →

Demos and videos

About The Apache Software Foundation

Polaris is an Apache Software Foundation top-level project, having graduated from the incubator in February 2026 after entering it in August 2024. The code was written at Snowflake, open-sourced in mid-2024 and granted to the ASF, and is Apache 2.0 licensed. Initial committers came from Snowflake, Dremio, Google, Microsoft, AWS and Confluent, and it is now governed by its own project management committee, chaired by Jean-Baptiste Onofré. There is no managed service from the foundation.

polaris.apache.org

Other table & technical catalogues tools

AWS Glue Data Catalog

Metadata Plane · Table & Technical Catalogues

Managed technical metadata catalogue for data on AWS, used by services such as Athena, EMR and Redshift.

  • Cloud service

Hive Metastore

Metadata Plane · Table & Technical Catalogues

Apache Hive's metadata service, still widely used as a table catalogue by Spark and other engines.

  • Open source

Unity Catalog

Metadata Plane · Table & Technical Catalogues

Governance catalogue for data and AI assets, available as open source and managed in Databricks.

  • Open core

Drafted with AI assistance and checked against the vendor’s own documentation.