Connecting Databricks Delta to Snowflake as Iceberg: Two Approaches — UniForm vs Delta Direct
Introduction
In today’s data-driven world, seamless integration between different platforms is crucial. Organizations are having multiple platforms in their tech landscape and interoperability is becoming inevitable. Some organizations have both Databricks and Snowflake in their data architecture and thats where we see excessive need to make the platforms tightly integrated and work seamlessly.
Two emerging approaches address this challenge:
- Databricks UniForm — Enables native Iceberg compatibility without additional metadata conversion.
- Snowflake Delta Direct — Converts Delta logs into Iceberg metadata for query execution in Snowflake.
This article explores their technical differences, use cases, and trade-offs.
Databricks UniForm: Unifying Data Formats
Introduced in Delta Lake 3.0 , UniForm is Databricks’ answer to the challenge of data interoperability. Delta UniForm enables seamless interoperability between Delta Lake and other open table formats like Apache Iceberg and Apache Hudi.
Key Features:
- Multi-format Metadata Generation: UniForm generates metadata for multiple table formats (Iceberg, Hudi) alongside Delta.
- Single Data Copy: Enables a single copy of data to be accessed by different ecosystems.
- Iceberg REST Catalog API: Implements this API for external access, facilitating integration with systems like Snowflake.
- Unity Catalog Integration: Supports Databricks’ unified governance solution.
By enabling Delta UniForm, organizations can leverage the advantages of multiple ecosystems while using Delta Lake, promoting flexibility and interoperability in data lakehouse architectures
When Does UniForm Shine?
UniForm is best suited for organizations that:
- Have a Multi-Engine or Multi-Cloud Strategy: If the same data needs to be queried by Databricks, Snowflake, Trino, or other Iceberg-compatible engines, UniForm ensures seamless access.
- Want to Maintain Delta Functionality: Features like Change Data Feed (CDC) and Time Travel remain fully supported.
- Prefer Databricks as the Primary Processing Engine: If data engineering, ML, or streaming workloads run on Databricks, UniForm allows external access without duplicating datasets.
Snowflake Delta Direct: Streamlined Delta Lake Access
Snowflake’s Delta Direct, currently in public preview offers a different approach to accessing Delta Lake data be it inDatabricks or not. Snowflake Delta Direct is a feature that allows users to access Delta Lake tables as Iceberg tables within Snowflake, without the need for data conversion or duplication
Key Features:
- Direct Delta Lake Access: Allows Snowflake to query Delta Lake tables without data conversion or duplication.
- Iceberg Table Creation: Creates Iceberg tables in Snowflake that point to Delta files.
- No UniForm Requirement: Does not require UniForm enablement on Delta tables.
- Optimized for Snowflake Workloads: Snowflake users can leverage Snowflake’s compute and governance model to process Delta data natively.
Delta Direct aims to provide a more streamlined approach for organizations looking to bridge their Delta Lake and Snowflake ecosystems, offering improved data accessibility and reduced complexity in managing data across these platforms
When Does Delta Direct Shine?
Delta Direct is ideal for organizations that:
- Are Primarily Snowflake-Driven: If analytics, BI, and governance are centered in Snowflake, Delta Direct provides native access without depending on Databricks.
- Need a Simplified Integration: There is no need to enable UniForm, modify metadata, or involve Databricks for configuration.
- Do Not Require Full Delta Functionality: Delta Direct does not support Change Data Feed (CDC), deletion vectors, or Time Travel but is optimized for read-heavy use cases.
Databricks UniForm
- Delta tables exist in Databricks with UniForm enabled.
- External compute engines like Snowflake, Trino, or Flink can read them as Iceberg or Hudi tables.
- Unity Catalog ensures governance consistency.
Snowflake Delta Direct
- Delta tables reside on cloud storage (S3/GCS/Azure).
- Snowflake interprets Delta metadata as Iceberg on the fly.
- No changes required on the Databricks side
Conclusion
Both UniForm and Delta Direct enable interoperability between Databricks and Snowflake, but they serve different needs:
- If Databricks is your main processing engine and you need multi-engine support → Use UniForm. It’s the best long-term approach since it removes metadata sync overhead.
- If Snowflake is your main analytics platform and you just need direct Delta access → Use Delta Direct but comes with additional metadata management costs.
As these technologies mature, organizations should assess their specific requirements — performance, governance, and ecosystem strategy — to select the best approach for seamless data integration.
