Snowflake Horizon Enhances Iceberg Interoperability Against Databricks Unity Catalog
Snowflake's Horizon Catalog now fully implements the Iceberg REST Catalog protocol, enabling bidirectional read/write access and secure credential vending for external Iceberg tables. This contrasts with Databricks Unity Catalog, which offers only inbound access and relies on less secure, long-lived credentials for outbound operations. This update matters for organizations seeking true data agency and reduced vendor lock-in, particularly impacting data engineers and architects managing multi-cloud or hybrid data platforms.
- →Snowflake Horizon's Full Iceberg REST Catalog Implementation
- →Vended Credentials for Enhanced Security in Iceberg
- →Defining Open Interoperability in Data Platforms
- →Databricks Unity Catalog's Interoperability Limitations Highlighted
- →Scan Plan API Extends Fine-Grained Governance
Features (2) ›
- Snowflake Horizon's Full Iceberg REST Catalog Implementation
Snowflake Horizon Catalog provides a complete implementation of the Iceberg REST Catalog protocol, supporting bidirectional federation for reading and writing to Iceberg tables. It also features vended credentials for secure, short-lived token delegation.
- Vended Credentials for Enhanced Security in Iceberg
The Iceberg REST Catalog specification includes vended credentials for secure, short-lived tokens, delegating access from the source catalog. Snowflake Horizon fully implements this for both inbound and outbound federation, while Unity Catalog does not, relying instead on broader IAM roles and long-lived credentials.
Enhancements (2) ›
- Databricks Unity Catalog's Interoperability Limitations Highlighted
The article asserts that Databricks Unity Catalog only supports inbound federation for Iceberg tables and relies on the Snowflake Catalog SDK with JDBC for outbound access, which is read-only. It also states Unity Catalog does not use Iceberg REST Catalog APIs or vended credentials for external tables.
- Scan Plan API Extends Fine-Grained Governance
Snowflake's Scan Plan API, currently in private preview, extends governance to fine-grained row and column level access controls for compatible engines. This feature builds upon existing governance capabilities within the Horizon Catalog.
Notes (2) ›
- Defining Open Interoperability in Data Platforms
Open interoperability allows secure access to governed data via open, standards-based interfaces, reducing proprietary lock-in and improving ecosystem compatibility. Snowflake emphasizes this principle in its support for Apache Iceberg and Apache Polaris, enabling customers to platform Iceberg tables on or off Snowflake without catalog migration.
- Databricks Compute Without Unity Catalog
Customers can still leverage Databricks compute with full read/write capabilities to any Iceberg REST Catalog-compliant catalog by configuring a Spark cluster without Unity Catalog enabled.
https://www.snowflake.com/content/snowflake-site/global/en/blog/bidirectional-interoperability-snowflake-horizon-databricks
Related releases
- Snowflake Powers Team USA Bobsled/Skeleton Data Collaboration Snowflake Blog ·
- Open Semantic Interchange Renamed Apache Ossie (Incubating) Snowflake Blog ·
- Marketers Need to Own Their AI Context Layer for Differentiation Snowflake Blog ·
- Snowflake Advocates Owning the Context Layer for AI Snowflake Blog ·
- Snowflake's Snowplan: AI-powered financial planning platform Snowflake Blog ·
- Snowflake Data Clean Rooms ML Jobs Now Generally Available Snowflake Blog ·