Delta Lake is an open-source table format (storage layer) for cloud data lakes with ACID transactions, time travel and many more (you'd rather not wanna miss in your data-heavy architectures).
Delta Lake allows you to store data on blob stores like HDFS, S3, Azure Data Lake, GCS, query from many processing engines including Apache Spark, Trino, Apache Hive, Apache Flink, and provides APIs for Scala, Java, Python, Rust, and Ruby.
As it was well said: "Delta is a storage format while Spark is an execution engine...to separate storage from compute." Yet, Delta Lake can run with other execution engines like Trino or Apache Flink.
Delta Lake 2.3.0rc1 supports Apache Spark 3.3.2 (cf. build.sbt).
Delta tables are parquet tables with a transactional log.
Changes to (the state of) a delta table are reflected as actions and persisted to the transactional log (in JSON format).
Delta Lake uses OptimisticTransaction for transactional writes. A commit is successful when the transaction can write the actions to a delta file (in the transactional log). In case the delta file for the commit version already exists, the transaction is retried.
More importantly, multiple queries can write to the same delta table simultaneously (at exactly the same time).
TransactionalWrite is an interface for writing out data to a delta table.
The following commands and operations can transactionally write new data files out to a data directory of a delta table:
Delta Lake provides the following Developer APIs for developers to interact with (and even extend) Delta Lake using a supported programming language:
Delta Lake supports batch and streaming queries (Spark SQL and Structured Streaming, respectively) using delta format.
In order to fine tune queries over data in Delta Lake use options.
Structured queries can write (transactionally) to a delta table using the following interfaces:
- WriteIntoDelta command for batch queries (Spark SQL)
- DeltaSink for streaming queries (Spark Structured Streaming)
Delta Lake supports reading and writing in batch queries:
Batch reads (as a
Batch writes (as a
Delta Lake supports reading and writing in streaming queries:
Stream reads (as a
Stream writes (as a
Delta Lake uses LogStore abstraction for reading and writing physical log files and checkpoints (using Hadoop FileSystem API).
Delta Tables in Logical Query Plans¶
Delta Table defines
DeltaTable Scala extractor to find delta tables in a logical query plan. The extractor finds
LogicalRelations (Spark SQL) with
HadoopFsRelation (Spark SQL) and TahoeFileIndex.
Put simply, delta tables are
HadoopFsRelation with TahoeFileIndex in logical query plans.
Concurrent Blind Append Transactions¶
A transaction can be blind append when simply appends new data to a table with no reliance on existing data (and without reading or modifying it).
Blind append transactions are marked in the commit info to distinguish them from read-modify-appends (deletes, merges or updates) and assume no conflict between concurrent transactions.
Blind Append Transactions allow for concurrent updates.
df.format("delta") .mode("append") .save(...)
Delta Lake supports Generated Columns.
Delta Lake introduces table constraints to ensure data quality and integrity (during writes).
Exception Public API¶
Delta Lake introduces exceptions due to conflicts between concurrent operations as a public API.
Simplified Storage Configuration¶
Delta Lake 1.2.0¶
Compacting Small Files (Optimize)¶
Delta Lake 1.2.0 introduces a new OPTIMIZE SQL command for compacting small files into larger ones.
Delta Lake 1.2.0 introduces support for Data Skipping.