Access on-premises data using OpenSharing
OpenSharing is an open protocol for secure, governed sharing of data and AI assets. It evolves the Delta Sharing protocol, and it lets you share data and AI assets across any cloud and any vendor. OpenSharing is also the basis for Databricks Marketplace.
With OpenSharing, you can share data assets, such as tables from MySQL, PostgreSQL, Snowflake, Amazon Redshift, Azure SQL, Azure Synapse, and Google BigQuery, and AI assets, such as Genie Agents, AI models, and AI skills. Recipients can be customers, partners, or any compatible client.
What problem it solves
OpenSharing provides a standard protocol for sharing data and AI assets, including agent skills and AI models, across organizations. Data stays in your own storage, with zero copies and no data movement.
OpenSharing uses a read and share pattern with standard APIs for discovery, authorization, and access, regardless of platform.
When to use OpenSharing
Use OpenSharing when you want to limit data duplication and query the freshest possible data. Otherwise, Databricks recommends ingesting data using Lakeflow Connect managed connectors.
Network requirements
OpenSharing does not require Cloud Interconnect or a site-to-site VPN. It does require the following:
- Network reachability to the OpenSharing endpoint and to the underlying object storage.
- TLS and authentication.
- Firewall rules that allow the traffic.
OpenSharing SecureConnect provides a Databricks-managed, simplified connectivity option. See Share data behind a firewall with SecureConnect.
Considerations
To give Databricks governed access to on-premises data with OpenSharing:
- Store your on-premises data in object storage, preferably an S3-compatible platform, and organize it as Delta or Iceberg tables.
- Make sure your storage platform exposes a compatible OpenSharing service or endpoint. Storage partners such as Everpure, MinIO, and Qumulo typically provide this endpoint.
- Connect the OpenSharing endpoint to Unity Catalog.