Skip to main content
Delta Sharing is an open protocol for secure real-time exchange of large datasets, enabling organizations to share data across different computing platforms without complex integrations or data duplication.

​
What is Delta Sharing?

Delta Sharing is a simple REST protocol that securely shares access to part of a cloud dataset. It leverages modern cloud storage systems—such as S3, ADLS, or GCS—to reliably transfer large datasets.
With Delta Sharing, users can connect directly to shared data through pandas, Tableau, Apache Spark, or any system supporting the protocol, without deploying specific compute platforms first.

​
Key Benefits

Universal Access

Share data once and reach consumers using Python, Spark, Tableau, Power BI, and more

No Data Duplication

Recipients access data directly from cloud storage without copying

Real-Time Updates

Consumers always see the latest data as providers update tables

Simple Integration

Minutes to start consuming data, not months of platform deployment

​
Core Concepts

Delta Sharing organizes shared data using a three-level hierarchy:

​
Share

A share is a logical grouping of data shared with recipients. A share can be shared with one or multiple recipients, who can access all resources within it. A share may contain multiple schemas. 
vaccine_share/
├── acme_vaccine_ingredient_data/
└── acme_vaccine_patient_data/

​
Schema

A schema is a logical grouping of tables within a share. Schemas help organize related tables together. 
vaccine_share/
└── acme_vaccine_data/
    ├── vaccine_ingredients
    ├── vaccine_patients
    └── vaccine_trials

​
Table

A table is a Delta Lake table or a view on top of a Delta Lake table. Tables contain the actual data that recipients query.
A table is referenced using the format:
<share-name>.<schema-name>.<table-name>
For example:
vaccine_share.acme_vaccine_data.vaccine_patients

​
Recipient

A recipient is a principal with a bearer token to access shared tables. Recipients authenticate using profile files containing their credentials.

​
How Delta Sharing Works

1

Provider Shares Data

The data provider configures a Delta Sharing Server and creates shares containing schemas and tables. They generate profile files for recipients.
2

Recipient Gets Profile

Recipients receive a profile file (JSON) containing:
  • Server endpoint URL
  • Authentication token
  • Protocol version
3

Client Connects

Recipients use Delta Sharing connectors (Python, Spark, etc.) with the profile file to connect to the sharing server.
4

Data Access

The connector queries the REST API to:
  • List available tables
  • Get table metadata and schema
  • Receive pre-signed URLs or credentials to read data files directly from cloud storage

​
Architecture Overview

Delta Sharing uses a simple architecture that separates metadata management from data transfer:
Delta Sharing Server
  • Authenticates recipients
  • Provides table metadata and schemas
  • Issues pre-signed URLs or temporary credentials for data access
  • Does NOT serve the actual data
Cloud Storage
  • Hosts the actual Parquet data files
  • Supports S3, ADLS, GCS, and more
  • Recipients read directly using pre-signed URLs
Delta Sharing Client
  • Python Connector: Read as pandas DataFrames
  • Spark Connector: Read as Spark DataFrames
  • Community connectors: Power BI, Tableau, R, Rust, Java, and more

​
Access Modes

Delta Sharing supports two ways to access table data:

​
URL-Based Access

The server returns pre-signed URLs for individual data files. The client fetches files via HTTP.
  • Best for: Ad-hoc queries, small-to-medium datasets
  • Advantages: Simple, works everywhere
  • How it works: Server generates time-limited URLs for each Parquet file

​
Directory-Based Access

The server issues temporary cloud credentials (e.g., AWS STS tokens) so clients can read the Delta log and data files directly using cloud storage APIs.
  • Best for: Large datasets, streaming, advanced Delta features
  • Advantages: Better performance, supports Change Data Feed (CDF)
  • How it works: Client reads Delta transaction log and data files directly
Most servers support both access modes. The client automatically chooses the best method based on capabilities.

​
Supported Platforms

Delta Sharing has official connectors and a growing ecosystem of community implementations:

​
Official Connectors

  • Python Connector: Read shared tables as pandas or Spark DataFrames
  • Apache Spark Connector: Read shared tables in Spark SQL, PySpark, Scala, Java, or R
  • Delta Sharing Server: Reference implementation for sharing Delta Lake and Parquet tables

​
Community Connectors

  • Power BI (Databricks)
  • Tableau
  • R, Rust, Java, Node.js, Clojure
  • Google Sheets
  • And many more…

​
What’s Next?

Ready to start using Delta Sharing? Continue to the Quickstart to get up and running in minutes.

Quickstart

Get started with Delta Sharing in minutes

Protocol Specification

Dive deep into the Delta Sharing Protocol

Build docs developers (and LLMs) love

Get started for freeTalk to us