Momentum One Academy · glossary
Delta Lake
Definition
Delta Lake is an open-source storage layer that combines analytical Parquet files with a transaction log. The log records committed changes, allowing ACID properties—atomicity, consistency, isolation, and durability—over data-lake files. It also supports capabilities such as schema management and time travel, so a table can be read as a known historical snapshot. 1
Where it sits in Fabric
Delta Lake is Fabric’s default table format in a lakehouse and its common storage format across Fabric workloads. A lakehouse exposes managed Delta tables in Tables, while Files is used for unstructured or non-Delta content. Spark notebooks and Spark SQL can create, transform, and query the tables; the lakehouse SQL analytics endpoint and Power BI can consume the same Delta data without a separate copy. 2 3 Fabric’s experiences do not all support every Delta feature, so interoperability depends on the writer, reader, runtime, and table features in use. 4
When to use it
Use Delta Lake for curated or actively transformed data that must remain reliable across ingestion, engineering, SQL analysis, and BI. It is especially useful when concurrent writes, schema changes, reproducible analysis, or historical inspection matter. For a simple landing area of files that are not yet modeled as tables, keep the data in the lakehouse’s Files area until a deliberate table-conversion step.
Practical example
A notebook can publish a curated sales table, then use Spark SQL to investigate a pipeline change without modifying today’s data:
DESCRIBE HISTORY sales;
SELECT * FROM sales VERSION AS OF 5;
DESCRIBE HISTORY helps identify an available version; the time-travel query reads that snapshot only. 5
Production caution
Time travel is not a backup or an unlimited audit archive. It needs both the relevant transaction-log entries and the historical data files. Log retention and VACUUM cleanup therefore determine how far back a version remains queryable; removing unreferenced files can make an older snapshot unavailable. Set retention and maintenance policies against the organization’s recovery, audit, and reproducibility requirements before shortening cleanup windows. 5
Workshop connections
In Module 2, place Delta Lake in the Lakehouse and OneLake storage model. In Module 4, use Spark SQL to create and transform Delta tables. In Module 5, use table history and Time Travel to validate pipeline results. In Module 6, discuss how shared Delta tables move between Spark, SQL, and Power BI. In Module 7, apply interoperability checks and production maintenance decisions.
Ready to turn these patterns into a governed delivery? Explore the Momentum One Academy Fabric Production Workshop.
References
Plan your workshop