|Databricks Delta Lake Now Open Source|
|Written by Kay Ewbank|
|Friday, 26 April 2019|
At the Spark +AI Summit taking place this week in San Francisco, Databricks announced that it has open sourced its Delta Lake storage layer, which handles the stage where data is brought into an organization's data lake.
Databricks was created as a company by the original developers of Apache Spark and specializes in commercial technologies that make use of Spark. Until now, Delta Lake has been part of Databricks Delta, the proprietary stack from Databricks. It is a unified analytics engine and associated table format built on top of Apache Spark.
Delta Lake is a storage layer that stores data in Apache Parquet format. It is designed for use in data lakes that are built on HDFS and cloud storage.
Data lakes are used to store both structured and unstructured data, but the data can be unreliable because of problems including schema mismatches and no enforcing of consistency. Data can be missing from some columns, and inconsistencies can creep in when schemas are changed in some parts of a pipeline but not in others.
Databricks Delta keeps closer control over the schemas in different parts of the data lake, validating that schema changes are replicated throughout the pipeline. Missing columns of data are correctly set to null, and data definition language (DDL) is used to add new columns and update schemas.These features and the use of optimistic concurrency control between writes, and snapshot isolation for consistent reads during writes, mean that Delta Lake offers ACID transaction support. Delta Lake also uses snapshots to give data versioning for rollbacks and reproducing reports. The tool has options such as schema enforcement, and all data in Delta Lake is stored in Apache Parquet format, a favorite format for storing and working with large datasets.
Another advantage Delta Lake offers is that you can carry out local development and debugging to develop data pipelines on your desktop or laptop machine. Delta Lake uses the Spark engine for the metadata of the data lake, and is compatible with the Apache Spark APIs.
Databricks says Delta is 10 -100 times faster than Apache Spark on Parquet. It has been designed for both batch and stream processing, and can be used for pipeline development, data management, and query serving.
Now that Delta Lake is open source, Databricks is open to contributions from outside the company.
or email your comment to: firstname.lastname@example.org
|Last Updated ( Friday, 26 April 2019 )|