Topic 3 of 6
Microsoft Fabric and Azure Databricks
Azure offers two main platforms for large-scale analytics. Both cover the full analytics architecture, but with different approaches: Fabric is a unified SaaS platform and Databricks is a code-first platform built on Apache Spark. The exam asks you to describe them and recognize when each one fits.
Contents
Two analytics platforms
Microsoft Fabric
A unified SaaS analytics platform. It brings data engineering, warehousing, real-time analytics, data science, and Power BI together in a single browser-based workspace on top of shared storage, OneLake. You don't manage servers or clusters.
Azure Databricks
A cloud analytics platform built on Apache Spark, optimized for data engineering, data science, and SQL analytics over open lakehouse formats. A managed Azure service, common for code-first, notebook-based workflows.
A third service, Microsoft Purview, focuses on data governance across all sources: it builds a map of the data and tracks its lineage.
Microsoft Fabric
An end-to-end analytics platform: ingestion, transformation, real-time processing, analytics, and reporting. It's delivered as SaaS: you create workspaces and items, and Microsoft runs the infrastructure.
- Role-specific workloads (engineers, data scientists, analysts, database administrators) that combine into a single solution.
- Integration with Microsoft 365 (Excel, Teams) and with Microsoft Foundry for AI models.
- It implements a data mesh architecture.
Fabric workloads
Data Factory
Data integration with Power Query and more than 200 connectors.
Data Engineering
Apache Spark with notebooks and jobs to process large volumes; includes the Lakehouse.
Data Warehouse
SQL performance that separates compute and storage and stores data in Delta Lake.
Data Science
Build, deploy, and operationalize machine learning models.
Real-Time Intelligence
Analyze data as it arrives: IoT sensors, logs, web clickstreams.
Databases
Transactional databases in Fabric (SQL database and Cosmos DB in Fabric) and mirroring of other systems.
Power BI
Connect to data, build interactive reports and dashboards, and share them.
Fabric IQ (preview)
Unifies business semantics so people, tools, and AI agents share definitions like Customer or Order.
OneLake
A unified data lake for the whole organization. Every Fabric tenant automatically includes a single OneLake: you can't delete it or create several, and there's no infrastructure to provision.
- It's built on Azure Data Lake Storage and stores tables in open formats: Delta Parquet or Iceberg.
- One copy of data: every Fabric engine (T-SQL, Spark, Analysis Services…) works on the same copy, without duplicating it.
- Hierarchy: tenant (shared policies), workspaces (each on a capacity tied to a region), and items (lakehouses, warehouses, eventhouses).
- OneLake security roles define permissions down to folders, tables, rows, and columns, enforced across every engine.
- It supports shortcuts and mirroring, the OneLake file explorer for Windows, and the ADLS Gen2 APIs.
Copilot and governance in Fabric
Copilot
AI built into the workloads to build pipelines, write SQL, generate notebook code, summarize, and explore data in natural language, respecting permissions.
Governance
The OneLake catalog centralizes discovery, administration, and governance of items. Microsoft Purview adds information protection, DLP, auditing, and compliance for supported data.
Azure Databricks
A unified, open analytics platform for building, deploying, and maintaining data, analytics, and AI solutions at scale. It runs as a managed service inside your Azure subscription and uses Delta Lake as its native storage format.
- Several open-source projects were created at Databricks: Delta Lake, MLflow, Apache Spark and Structured Streaming, and Unity Catalog.
- Use cases: enterprise lakehouse, ETL and data engineering, machine learning and AI, warehousing and BI.
- Engineers reuse their Spark skills, and analysts query and visualize data from browser-based notebooks.
Azure Databricks components
Databricks Lakehouse
A unified storage layer on Delta Lake for SQL analytics and machine learning in the same environment.
Databricks SQL
SQL warehouses (serverless, pro, or classic) for querying Delta tables, with query history, dashboards, and alerts.
Notebooks
Collaborative Python, SQL, Scala, and R notebooks for pipelines, exploratory analysis, and machine learning.
Unity Catalog
A unified governance layer for data and AI: fine-grained access control, lineage, and discovery.
Genie
An AI-powered conversational interface: users ask in natural language and Genie generates and runs the SQL.
Jobs
Schedule notebooks, SQL queries, and other code.
Fabric vs. Databricks
| Microsoft Fabric | Azure Databricks | |
|---|---|---|
| Model | Unified SaaS | A managed service in your subscription |
| Storage | OneLake (Delta Lake) | Delta Lake on your cloud storage |
| Style | Low-code to code, all in one workspace | Code-first: Spark and notebooks |
| Governance | OneLake catalog and Purview | Unity Catalog |
| AI | Copilot | Genie and AI assistance |
| Choose it if | You want an integrated solution with Power BI and no infrastructure to manage | You need code-first Spark workflows, multicloud portability, or already have Databricks expertise |
Exam scenarios
| Situation | Answer | Why |
|---|---|---|
| A SaaS platform where engineering, warehouse, real time, and Power BI share one store | Microsoft Fabric | Everything on OneLake, no infrastructure to manage. |
| A single data lake for the whole organization that can't be duplicated or deleted | OneLake | One per tenant. |
| Restrict a partner to only see rows where Region = "US" | OneLake security roles | Permissions down to rows and columns. |
| A team experienced with Spark that wants notebooks and multicloud portability | Azure Databricks | A code-first, Spark-based platform. |
| Unified data and AI governance with lineage in Databricks | Unity Catalog | Access control, lineage, and discovery. |
| Natural language questions about data in Databricks | Genie | It generates and runs SQL. |
| Write SQL and notebook code with AI help in Fabric | Copilot in Fabric | AI built into the workloads. |
| Unify business definitions like Customer or Order for people and AI agents | Fabric IQ (preview) | A shared ontology and semantics. |
Common exam traps
❓ Is Microsoft Fabric PaaS or SaaS?
SaaS: you don't manage servers or clusters.
❓ Can you create several OneLakes in a tenant?
No. Each tenant has a single OneLake, which can't be deleted.
❓ What is OneLake built on?
Azure Data Lake Storage, with tables in Delta Parquet or Iceberg.
❓ Which engine is Azure Databricks built on?
Apache Spark.
❓ Which Databricks component provides governance, lineage, and access control?
Unity Catalog.
❓ Do you need to copy data to use it with both Spark and T-SQL in Fabric?
No. Every engine works on the same copy in OneLake.
❓ Which AI assistant does Fabric have, and which does Databricks have?
Copilot in Fabric; Genie (and other AI features) in Databricks.
Did you get this topic?
Test what you just learned
An organization wants a single platform where data engineers, analysts, and data scientists work on one copy of the data, with Power BI built in and no clusters to manage. What does it choose?
Ready to test "Microsoft Fabric and Azure Databricks"?
Original questions, each with a full explanation.
Start a mock examSign in to save your progress.