Lakehouse vs. warehouse for big data

Introduction to Microsoft Fabric

Fiodar Sazanavets

Senior Software Engineer

Why Fabric has different data stores

Fabric data stores split into a lakehouse workshop of raw materials and parts and a warehouse of labeled boxes on shelves

  • Different work needs different storage
  • Lakehouses: raw and structured data
  • Warehouses store structured data
Introduction to Microsoft Fabric

What is a lakehouse?

Lakehouse as a two-level workshop where raw CSV and JSON files upstairs become Delta tables on shelves below

  • Like a flexible workshop
  • Raw files (CSV, JSON, Parquet) and tables
  • Useful for data engineering and data science
Introduction to Microsoft Fabric

What is a warehouse?

  • Like a managed stockroom
  • Structured tables only
  • Designed for SQL-first analytics
  • Familiar to analysts and database users

Warehouse as a stockroom with a schema and labeled shelves of Revenue, Costs, Customers, and Accounts tables

Introduction to Microsoft Fabric

Common lakehouse scenarios

Raw CSV and JSON files in crates pass through a workbench to become lakehouse tables

  • Data arrives as files; originals kept
  • Transform files into tables
Introduction to Microsoft Fabric

Common warehouse scenarios

  • Store curated business records
  • Use a consistent structure

Structured data flows into a Fabric warehouse of fixed-schema tables that serves reporting, analytics, finance, and governance

Introduction to Microsoft Fabric

The role of SQL for downstream analytics

  • Downstream analytics: work after data is stored
    • Querying, modeling, and reporting
  • Analyses need only a slice of the data
  • SQL defines which data to pull
Introduction to Microsoft Fabric

How downstream analytics works

  • Data preparation layer: curated tables
  • Modeling layer: business metrics and relationships
  • Reporting layer: what end users read

Prepared data from a lakehouse and a warehouse feeds a semantic model, which then feeds a report

Introduction to Microsoft Fabric

Lakehouse vs. warehouse

Lakehouse

  • Stores raw files and tables
  • Read-only SQL analytics endpoint

Warehouse

  • Only stores tables
  • Full T-SQL, including INSERT and UPDATE
Introduction to Microsoft Fabric

Let's practice!

Introduction to Microsoft Fabric

Preparing Video For Download...