The Challenge of Traditional Data Sharing
Sharing data across organizational boundaries has traditionally been a complex and inefficient process. Common approaches involve exporting datasets as CSV files, compressing them, transferring them through email or FTP, and relying on recipients to work with static snapshots. By the time those files are accessed, they may already be outdated. This process introduces unnecessary manual effort, duplicate data copies, and significant security and governance challenges.
Delta Sharing was introduced to address these limitations by enabling organizations to share live data securely without copying or moving it. This article explores what Delta Sharing is, why it matters, and how to configure it in Databricks using real-world examples and screenshots.
What is Delta Sharing?
Delta Sharing is an open protocol developed by Databricks and open-sourced under the Linux Foundation that enables organizations to share live data with internal teams, customers, and external partners without physically copying or moving it. The data remains in its original cloud storage location (Amazon S3, Azure Data Lake Storage, or Google Cloud Storage), while recipients receive secure, read-only access.
One of Delta Sharing’s biggest advantages is its open ecosystem. Recipients do not need a Databricks workspace to consume shared data. They can access it using tools such as pandas, Apache Spark, Power BI, Tableau, or any client that supports the open Delta Sharing protocol. This makes cross-platform and cross-organization collaboration significantly simpler.
Behind the scenes, when a recipient queries shared data, Databricks generates short-lived pre-signed URLs that point directly to the underlying cloud storage files. For shared tables, recipients read data directly from storage, eliminating the need for provider-side compute and making the solution highly scalable and cost-efficient. Access is centrally governed through Unity Catalog, allowing administrators to grant or revoke permissions at any time.
For Databricks-to-Databricks sharing, authentication is handled natively through Unity Catalog without exchanging credentials. For non-Databricks platforms, secure authentication is supported through short-lived access tokens or OIDC federation.
Why Choose Delta Sharing?
One of the primary advantages of Delta Sharing is data freshness. Traditional file-sharing methods distribute static snapshots that quickly become outdated, requiring repeated exports and synchronization. Delta Sharing eliminates this challenge by providing recipients with access to the latest version of the data whenever they query it.
Some of the key benefits include:
Zero data duplication: Data remains in its original cloud storage location, eliminating the need to create and manage multiple copies.
Live data access: Updates made to shared tables become immediately available to recipients without scheduled refreshes or synchronization jobs.
Cross-platform compatibility: Recipients can use their preferred analytics or BI tools, regardless of whether they use Databricks.
Open protocol: External users do not require a Databricks license, making collaboration with customers and partners straightforward.
Built-in governance: Unity Catalog provides centralized access control, audit logging, IP access restrictions, and token management for secure data sharing.
Time Travel support: Historical versions of shared tables can be exposed when required for auditing, compliance, or reproducibility.
End-to-end security: Short-lived pre-signed URLs minimize security risks, while revoking recipient access immediately prevents new requests, with existing URLs expiring automatically within minutes.
Traditional Data Sharing vs. Delta Sharing
The following comparison highlights the differences between conventional data-sharing approaches and Delta Sharing.
Delta Sharing | CSV / Email | ETL Pipeline | Custom API | |
Data freshness | Live / Near real-time | Stale | Scheduled | Depends |
Data copies | Zero | Full copy | Full copy | Full copy |
Setup effort | Low | Very low | High | Very high |
Governance | Unity Catalog | Manual | Custom | Limited |
Cross-org access | Seamless | Manual | Complex | Complex |
Recipient tool | Any client | Any | Depends | Custom |
When Does Delta Sharing Make Sense?
Delta Sharing is designed for a wide range of data collaboration scenarios where secure, governed, and real-time access to data is essential. Some of the most common use cases include:
Sharing data with external partners: Suppliers, vendors, or business partners often require access to operational data such as inventory, sales, or order information. Instead of relying on scheduled exports, Delta Sharing provides secure, live access to the latest data.
Customer-facing data delivery: Organizations offering SaaS products can securely provide customers with access to their own usage data, analytics, or reports directly from Databricks without creating duplicate datasets.
Cross-team collaboration within an enterprise: Large organizations frequently operate multiple Databricks workspaces across business units. Delta Sharing enables teams to securely exchange data without building centralized ETL pipelines or duplicating data.
Compliance and audit reporting: Regulatory bodies and auditors often require read-only access to specific datasets. Delta Sharing provides governed access with comprehensive audit logging while maintaining data security.
Building a data marketplace: Delta Sharing serves as the underlying protocol for Databricks Marketplace, enabling organizations to publish, discover, and distribute data products at scale.
Beyond Tables: What Can Be Shared?
Delta Sharing has evolved far beyond sharing structured tables. Today, it supports a broad range of governed data assets, enabling organizations to collaborate across multiple workloads.
Volumes: Share unstructured files such as PDFs, images, CSV files, JSON documents, and other assets stored in Databricks Volumes.
AI Models: Securely share machine learning models across teams or organizations without manually packaging and distributing model artifacts.
Streaming Tables: Allow recipients to consume shared data as a Structured Streaming source, enabling near real-time data pipelines.
Materialized Views: Share precomputed aggregations to deliver faster query performance for complex analytical workloads.
Databricks Marketplace: Publish and distribute data products through Databricks Marketplace using the Delta Sharing protocol.
Clean Rooms: Enable privacy-preserving analytics where multiple organizations collaborate on combined datasets without exposing their underlying raw data.
Change Data Feed (CDF): Allow downstream consumers to process only incremental changes instead of re-reading entire datasets, improving efficiency and reducing processing costs.
Delta Sharing has evolved into a comprehensive data collaboration platform, enabling secure sharing of structured data, unstructured files, AI models, streaming datasets, and other governed assets—all managed through Unity Catalog.
How Delta Sharing Works
Although the underlying architecture is sophisticated, the workflow is straightforward.
Delta Sharing involves two primary roles:
Provider – The organization or workspace that owns and shares the data.
Recipient – The user, team, or organization that receives access to the shared data.
The provider creates a Share, which acts as a logical container for the data assets to be exposed. These assets can include tables, views, volumes, AI models, or notebooks. One or more recipients are then granted access to the share.
The data flow follows this sequence:
Provider → Delta Sharing Server (Databricks) → Recipient
When a recipient queries shared data, Databricks validates permissions through Unity Catalog before generating short-lived pre-signed URLs that point directly to the underlying Delta Lake files stored in cloud storage. The recipient accesses the data directly from storage, ensuring zero-copy data sharing while allowing the provider to retain complete control over governance and permissions.
Access can be revoked instantly at any time. Since the generated URLs are short-lived, any active sessions automatically expire within minutes. Similarly, when additional assets are added to an existing share, they become immediately available to authorized recipients without requiring any changes on their side.
Authentication depends on the sharing model:
Databricks-to-Databricks Sharing: Authentication is managed natively through Unity Catalog without exchanging credentials.
Open Delta Sharing: External platforms can authenticate using OIDC federation or short-lived bearer tokens that support the open Delta Sharing REST protocol.
Setting Up Delta Sharing
The following walkthrough demonstrates how to configure Delta Sharing within a Databricks workspace.
Step 1 – Open Catalog Explorer
Sign in to your Databricks workspace and navigate to Catalog from the left navigation pane. This opens Catalog Explorer, where catalogs, schemas, tables, models, volumes, and sharing configurations are managed.
At the top of the Catalog Explorer, select the Delta Sharing tab to access the sharing interface and begin configuring data sharing.
Step 2 – Open Delta Sharing
Click Delta Sharing in the navigation bar to open the Delta Sharing interface. The page contains two tabs:
Shared with me: Displays data that external providers have shared with your organization. For first-time users, this section may initially be empty. External sharing must be enabled by a Databricks administrator at the account level—a one-time configuration that allows all workspaces within the account to share and receive data securely.
Shared by me: Serves as the provider workspace where shares are created, managed, and assigned to recipients. This is the primary area for configuring and managing outbound data sharing.
The following steps focus on creating and managing data shares from the Shared by me tab.
Step 3 – Explore the Provider Dashboard
Switch to the Shared by me tab. This serves as the central dashboard for managing outbound data sharing. From here, shares, recipients, and sharing configurations can be created and managed.
Key components of the dashboard include:
Shares: Displays all created shares along with details such as the share name, creation date, and owner.
Recipients: Lists the organizations or users who have been granted access. Recipients can be managed independently of individual shares.
Sharing ID: A unique identifier for the Databricks workspace. When another Databricks workspace requests access, its Sharing ID is used to create a recipient for Databricks-to-Databricks sharing.
New Recipient: Creates a new recipient object, which can be configured before or during the share creation process.
Share Data: Launches the share creation wizard to configure and publish a new data share.
Tip: The Sharing ID is safe to share with external organizations. It serves only as a unique workspace identifier and does not grant access to any data on its own.
Step 4 – Create a Share
Click Share data to launch the share creation wizard. The wizard consists of four steps:
Create the share
Add data assets
Add notebooks (optional)
Add recipients
The first step requires a share name and an optional description. It is recommended to follow clear and consistent naming conventions, such as partner_inventory_feed or sales_data_q1_2025, to simplify share management as the number of shared assets grows across the organization.
If an orange banner indicating that external sharing is not enabled appears, a Databricks administrator must enable External Sharing at the account level. This is a one-time configuration that allows all workspaces within the account to securely share data with external organizations.
Step 5 – Select the Tables to Share
The second step of the wizard is where the data assets to be shared are selected. The left pane displays the complete catalog hierarchy, including catalogs, schemas, and tables. Expand the desired catalog and schema, then select the tables to include in the share.
In this example, the employees_dataset table from demo_catalog > delta_share_demo is selected. Once a table is chosen, its configuration options appear in the right pane, where the following settings can be configured:
Alias: Assigns an alternate name to the shared table that recipients will see. This is useful for presenting business-friendly names or preventing exposure of internal naming conventions.
History: Enabled by default, allowing recipients to use Delta Lake Time Travel to query previous versions of the shared table. Disable this option if only the latest version of the data should be accessible.
Partition Filtering: Restricts access to a subset of the table’s data based on specific partition values, such as region, country, or date. This provides fine-grained control over the data exposed to each recipient.
Tip: A single share can include tables from multiple catalogs and schemas. There is no requirement to create separate shares for each catalog, making it easier to organize and manage related datasets.
Step 6 – Add Notebooks (Optional)
The third step of the wizard allows notebooks to be included in the share. This can be particularly useful when sharing data with teams or partners who are unfamiliar with the dataset. A notebook can provide sample queries, data exploration examples, and documentation that helps recipients understand the structure and usage of the shared data.
To add a notebook, browse to the desired notebook, assign an alias if required, and select an appropriate storage location that aligns with the organization’s governance and security policies. Databricks centralizes notebook management, making it easy to maintain and distribute supporting resources alongside shared data.
Adding notebooks is entirely optional. If no supporting documentation or sample code is required, this step can be skipped and the setup can proceed directly to configuring recipients.
Step 7 – Add Recipients to the Share
The final step of the wizard is to grant access to recipients. Select the Select a recipient dropdown to view all recipient objects that have been configured in the workspace. Choose the appropriate recipient, then click Share data to complete the setup.
Once the share is created, the selected recipient receives read-only access to the tables and other assets included in the share. They can immediately begin accessing the shared data based on the permissions granted.
Tip: A single share can be assigned to multiple recipients, and each recipient can have access to multiple shares. This many-to-many relationship provides the flexibility to organize data sharing based on business units, partners, customers, or specific use cases.
Step 8 – Create a New Recipient
If the intended recipient is not already available in the list, click + Create new recipient. A dialog box opens where the recipient’s details can be configured.
The following fields are required:
Recipient Name: Specify a meaningful name for the recipient, such as the organization or team name. Following a consistent naming convention helps simplify recipient management as the number of shared connections grows.
Recipient Type – Databricks: Select this option when sharing with another Databricks workspace. Authentication is managed natively through Unity Catalog, eliminating the need to exchange credentials. The recipient’s Sharing Identifier is required to establish the connection.
Recipient Type – Open: Select this option when sharing with users or organizations that are not using Databricks. Open recipients can authenticate using OIDC federation, the recommended modern approach, or through short-lived access tokens supported by the open Delta Sharing REST protocol.
Sharing Identifier: For Databricks-to-Databricks sharing, enter the recipient workspace’s unique Sharing Identifier. The identifier typically follows a format similar to:
aws:us-east-2:27bef0da-a2a3-4431-b418-a1a3e8b4bb54
After the recipient is created, it becomes available for selection during the share creation process and can be reused across multiple data shares.
Step 9 – Accessing Shared Data as a Recipient
Once a share has been created and access has been granted, recipients can seamlessly access the shared data from their own Databricks workspace.
After signing in, navigate to Catalog Explorer, where a new section named Delta Shares Received appears in the left navigation pane. All shares that have been made available by external providers are listed in this section.
To begin using a shared dataset, select the desired share and create a catalog from it. Once the catalog is created, the shared tables become available alongside other catalogs in the workspace and can be queried just like any native Databricks table.
Recipients can immediately leverage the shared data using familiar Databricks capabilities, including:
Running SQL queries
Processing data with PySpark
Building dashboards and visualizations
Consuming shared tables as Structured Streaming sources
Training and deploying machine learning models
The experience is designed to be seamless, allowing shared assets to integrate naturally into existing analytics and data engineering workflows without requiring any changes to how users typically work.
In the example shown below, the received share appears under Delta Shares Received in the Catalog Explorer. The shared tables are displayed in the main panel, making them immediately available for exploration and querying after the catalog is created.
Governance and Security
One of the standout strengths of Delta Sharing is its governance model, which is built on Unity Catalog. Rather than relying on custom security implementations, Delta Sharing provides centralized governance, fine-grained access control, and comprehensive auditing to ensure secure data collaboration.
Key governance capabilities include:
Audit Logs: Every query executed against shared data is recorded, providing complete visibility into who accessed which datasets, when the access occurred, and from where.
IP Access Lists: Access to shared data can be restricted to approved IP ranges, making it ideal for organizations that require connections only from corporate networks or VPNs.
Token Rotation: For Open Delta Sharing, authentication tokens can be rotated periodically without disrupting recipient access, improving overall security.
OIDC Federation: Modern identity-based authentication enables recipients to authenticate through their existing identity providers without exchanging credential files.
Instant Access Revocation: Recipient access can be revoked immediately. Since Delta Sharing uses short-lived pre-signed URLs, any active sessions automatically expire within minutes, minimizing security risks.
This multi-layered security and governance model makes Delta Sharing a robust alternative to traditional file transfers, custom APIs, and manually managed data-sharing solutions.
What’s New in 2025–2026
Databricks continues to enhance Delta Sharing with new capabilities that expand its role as a comprehensive data collaboration platform.
Recent enhancements include:
OIDC Federation: Modern authentication using enterprise identity providers, eliminating credential file management and simplifying secure access.
Volume Sharing: Support for sharing unstructured data such as PDFs, images, CSV files, and other assets stored in Databricks Volumes.
AI Model Sharing: Governed sharing of machine learning models across teams and organizations without manual packaging or distribution.
SAP BDC Connector: Enterprise integration between SAP and Databricks using Delta Sharing as the transport layer, enabling secure access to ERP data without custom pipelines.
Databricks Marketplace Expansion: Continued growth of the Marketplace ecosystem, allowing organizations to discover, subscribe to, and publish high-quality data products through Delta Sharing.
Best Practices
The following recommendations can help simplify administration while maintaining strong governance and security:
Use meaningful naming conventions: Consistent share names, such as
partner_inventory_feedorsales_partner_2025, make large-scale environments easier to manage.Share views instead of raw tables when appropriate: If a dataset contains sensitive information, create a view that exposes only the required columns before sharing it.
Disable history when unnecessary: If recipients only require the latest version of the data, disabling Time Travel reduces unnecessary storage usage and limits historical data exposure.
Adopt OIDC Federation for Open Sharing: Identity-based authentication eliminates the need to manage credential files and simplifies secure external collaboration.
Review audit logs regularly: Periodically monitoring Unity Catalog audit logs helps identify unexpected access patterns and supports compliance requirements.
What’s Next?
Delta Sharing serves as the foundation for several advanced Databricks capabilities. Once the fundamentals are in place, the following areas are worth exploring:
Databricks Marketplace: Discover, subscribe to, or publish third-party data products using the Delta Sharing protocol.
Clean Rooms: Enable privacy-preserving collaboration where multiple organizations can analyze combined datasets without exposing their raw data.
Data Mesh Architectures: Empower domain teams to publish governed data products while allowing other domains to securely consume them through Unity Catalog.
Streaming Collaboration: Build real-time data pipelines by consuming shared tables as Structured Streaming sources.
Conclusion
Delta Sharing simplifies one of the most challenging aspects of modern data platforms—secure collaboration across organizational boundaries. Its zero-copy architecture, open protocol governed by the Linux Foundation, and seamless integration with Unity Catalog enable organizations to share live data without duplicating datasets or building complex integration pipelines.
With continued enhancements such as Volume Sharing, AI Model Sharing, OIDC Federation, and deeper integration with Databricks Marketplace, Delta Sharing has evolved beyond simple table sharing into a comprehensive data collaboration platform.
Organizations that still rely on CSV exports, scheduled file transfers, or custom integration pipelines can significantly improve data freshness, security, governance, and operational efficiency by adopting Delta Sharing. With a straightforward setup process and centralized management through Unity Catalog, it provides a scalable and maintainable approach to secure data sharing across teams, partners, and customers.
