Data Mesh Governance / Policies / Interoperability / File Format
Delta File Format
Platform: Databricks, Azure Synapse Analytics, Generic Data Lake
Data products are stored as files on S3 (AWS S3 as Storage for Data Products).
To ensure interoperability and consistent usage patterns, we want to agree on a common file format.
We assume that data products frequently will be combined across domains.
We use Delta as file format for data products.
- Software engineers need to learn Delta and Parquet file format
- Delta is still not a wide-spread format beyond Databricks and Azure, limits use with other tools
- Low storage and IO costs through compression
- Fast querying and processing, as column-oriented file format
- Additional transaction log
- supports update and delete operations -> Do we want to use this feature for data products?
- Files may be fragmented after updates
- Time travel is possible
- Follow-Up Questions
- How to document the schema?
- Timestamp format
- How to model mutable data (like master and reference data)?
- Databricks comes with Delta support out of the box
- Automated testing: Query all data products periodically and try to deserialize latest file