Microsoft Fabric Lakehouse
The 10 Most Important Questions and Answers
What is a Lakehouse in Microsoft Fabric?
A Lakehouse is a single Microsoft Fabric item that combines the flexibility of a data lake with the SQL-oriented performance and management capabilities commonly associated with a data warehouse. It can store structured and unstructured data in OneLake, while managed tables are stored in the Delta Lake format.
What is OneLake, and how does the Lakehouse use it?
OneLake is Microsoft Fabric's unified organizational data lake. It is often described as being similar to OneDrive, but designed for enterprise data. Every Fabric Lakehouse stores its data in OneLake, helping organizations reduce unnecessary copies, connect different analytics workloads, and avoid isolated data silos.
Lakehouse vs. Warehouse: What is the key difference?
| Feature | Lakehouse | Warehouse |
|---|---|---|
| Primary engine | Apache Spark and SQL analytics endpoint | T-SQL relational engine |
| Data types | Structured, semi-structured, and unstructured | Primarily structured relational data |
| SQL endpoint | Primarily used for querying Lakehouse Delta tables | Supports read and write T-SQL operations |
| Common use | Data engineering, data science, and analytics | Enterprise reporting and relational warehousing |
What is the difference between the Tables and Files areas?
The Tables area contains managed Delta tables that can be queried through Spark and the SQL analytics endpoint. The Files area is commonly used as a landing or storage location for unmanaged files such as CSV, JSON, text, image, and Parquet files.
What is the SQL analytics endpoint?
A SQL analytics endpoint is automatically associated with a Lakehouse. It allows users to query Lakehouse Delta tables by using T-SQL and connect familiar SQL and reporting tools. It is commonly used for operations such as SELECT queries and creating SQL views. Data changes to Lakehouse tables are normally performed through Spark, pipelines, notebooks, or other supported ingestion and transformation methods.
What are OneLake Shortcuts?
OneLake Shortcuts are references to data stored in another supported location without requiring the data to be physically copied into the current Lakehouse. A shortcut can provide access to data in another Fabric item or an external storage system, creating a logical and unified data view while reducing duplication and additional storage requirements.
Why does the Lakehouse use the Delta Lake format?
Delta Lake extends the Parquet file format with capabilities such as ACID transactions, schema enforcement, schema evolution, and versioned data history. These capabilities provide greater reliability and make lake-based data easier to manage for analytics and warehouse-style workloads.
What is Direct Lake mode in Power BI?
Direct Lake is a Power BI semantic-model storage mode designed to work directly with data stored in OneLake. It aims to provide fast analytical performance without requiring the traditional full-data import process. This helps Power BI reports use Lakehouse or Warehouse data while reducing unnecessary data duplication.
What is the Medallion architecture?
Bronze
Raw source data stored with minimal transformation.
Silver
Cleaned, standardized, validated, and enriched data.
Gold
Business-ready data prepared for reporting and analytics.
How do you keep Lakehouse tables fast?
Lakehouse performance can be improved by reducing excessive small files, organizing table data appropriately, and applying supported optimization processes. Common techniques include V-Order optimization, OPTIMIZE for file compaction, and VACUUM for removing files that are no longer required after the applicable retention period. These practices can improve Spark, SQL, Direct Lake, and Power BI read performance.
Quick Summary
Microsoft Fabric Lakehouse brings together OneLake, Delta Lake, Apache Spark, SQL analytics, data engineering, data science, and Power BI. It enables organizations to store raw files and structured tables within a shared analytics environment while supporting layered data architecture, reusable data access, and modern business-intelligence workloads.