Unify analytics on one platform
Bring data integration, engineering, warehousing, real-time analytics and BI onto Microsoft Fabric over OneLake, use shortcuts and mirroring to avoid extra copies, and keep Azure Databricks where Spark-scale machine learning with Unity Catalog matters.
The problem
Many analytics estates grew one tool at a time: a pipeline service, a warehouse, a Spark cluster, a streaming store and a BI tool, each with its own copy of the data, its own security model and its own bill. Analysts wait for data to be moved, and governance teams cannot see who uses what.
Licensing adds confusion. Microsoft is retiring Power BI Premium per-capacity (P) SKUs and steers buyers to Fabric capacity, so BI and data platform decisions now affect each other.1
The approach
Microsoft Fabric is a SaaS analytics platform whose workloads (Data Factory, Data Engineering, Data Warehouse, Data Science, Real-Time Intelligence, Databases, Power BI and Fabric IQ, which is in preview) share one storage layer, OneLake. Each tenant gets one OneLake, built on Azure Data Lake Storage Gen2, and Microsoft says you do not need an Azure account to use Fabric.
- Avoid copies: shortcuts give zero-copy access to Azure Data Lake Storage, Amazon S3 and Google Cloud Storage; mirroring continuously replicates Azure SQL Database, Azure Cosmos DB, Azure Databricks and Snowflake data into OneLake.
- Govern centrally: OneLake Catalog for discovery and administration, with Microsoft Purview for information protection, DLP and audit on supported Fabric data.
- License with capacity: F SKUs are bought through Azure with per-second billing and no commitment, and Microsoft calls them the recommended capacities. On F64 or larger, users with a free license and a viewer role can view Power BI content; below F64 viewers need Pro or Premium Per User.
Fabric is not the only answer. Microsoft's architecture guide recommends Fabric for new analytics-integrated machine learning, Azure Databricks for lakehouse-scale Spark and machine learning with Unity Catalog governance, and keeping Azure Synapse Analytics only for existing investments, preferring Fabric for new ones. Mirroring Databricks data into OneLake lets the two coexist.
When a simpler route is enough: if the need is reports for a small team, Power BI Pro or Premium Per User may be enough. Microsoft notes that Premium Per User does not provision a Fabric capacity, so it cannot run lakehouses, warehouses or notebooks.123
Conceptual architecture
Diagram as a list
Copilot & business apps
- Power BIReports and dashboards for business users
Data & analytics
- Sources: databases, SaaS apps, files, streams, other cloudsOrigin of business data
- Data Factory pipelines and dataflowsIngest and transform data with many connectorsConnects to OneLake
- Shortcuts and mirroringZero-copy access and continuous replication into OneLakeConnects to OneLake
- OneLakeSingle logical data lake for all Fabric workloads
- Spark, Data Warehouse and Real-Time IntelligenceEngineering, SQL analytics and streaming analysisConnects to Power BI
Security, identity & governance
- OneLake Catalog and Microsoft PurviewDiscovery, access control, labels, DLP and auditConnects to OneLake
Cloud & hybrid infrastructure
- Fabric capacity (F SKU)Compute pool billed through AzureConnects to Spark, Data Warehouse and Real-Time Intelligence
Technologies and their roles
Microsoft Fabric2
Unified analytics platform
Workloads share OneLake, so teams reuse data without moving it between services.
Power BI1
Reporting
Power BI is a Fabric workload; capacity size decides whether free users can view content.
Azure Databricks3
Spark-scale data science
Microsoft recommends it for lakehouse-scale Spark and machine learning with Unity Catalog governance.
Microsoft Purview2
Governance
Adds information protection, DLP, audit and compliance controls for supported Fabric data.
Azure Synapse Analytics3
Existing estates
Keep for existing investments; Microsoft prefers Fabric for new ones and provides migration guidance.
What you need first1
- A Microsoft Entra tenant and a Fabric capacity (F SKU or trial) sized for the workloads
- At least one Power BI Pro or Premium Per User license to create and share Power BI content
- A domain and workspace plan that maps business areas to workspaces and owners
- An inventory of sources marked for shortcut, mirroring or pipeline ingestion
- Sensitivity labels and access rules agreed before data is shared widely
Risks and how to reduce them
- Viewers cannot open reports on small capacities1
- Below F64 every viewer needs Pro or Premium Per User; size capacity and licenses together.
- Assuming Premium Per User gives Fabric features1
- Premium Per User does not provision a Fabric capacity; add an F capacity for lakehouses, warehouses and notebooks.
- Duplicated data and cost from copying everything into Fabric2
- Use shortcuts and mirroring where they fit and copy only what needs transformation.
- Preview features in production plans2
- Fabric IQ is in preview; keep it out of critical paths until Microsoft announces general availability.
Documented examples
KPMG
ScalingKPMG: rebuilding the Digital Gateway data layer on Microsoft Fabric
about 2 hours instead of about 16 (87% less) (measured)
Professional services, Global (138 countries and territories)
Make-A-Wish America (national office and chapters)
In productionMake-A-Wish America: unified data, Copilot and an IT help bot
roughly two-thirds of staff (measured)
Nonprofit, United States
Related
Sources
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.