Migrating from Azure Synapse to Microsoft Fabric: A Practical Guide for UK Teams

Microsoft Fabric replaces Azure Synapse Analytics. We walk through the key differences, migration path, and what to watch out for.

Microsoft has made its strategic direction clear: Fabric is the future of the Azure data platform. Azure Synapse Analytics is not being retired imminently, but new capabilities are being built into Fabric first.

We cover the practical side of this in our Microsoft Fabric delivery, and rebuilding Wasabi's Microsoft estate looks at a closely related question.

Understanding what you're migrating from

• Synapse Dedicated SQL Pool - the managed data warehouse
• Synapse Serverless SQL - on-demand SQL queries over data lake files
• Synapse Spark Pools - managed Apache Spark for data engineering
• Synapse Pipelines - the orchestration layer, essentially Azure Data Factory embedded in Synapse

The Fabric equivalents

• Synapse Dedicated SQL Pool → Fabric Warehouse (T-SQL warehouse on OneLake)
• Synapse Serverless SQL → Fabric SQL Analytics Endpoint
• Synapse Spark Pools → Fabric Spark (same runtime, integrated into Fabric workspaces)
• Synapse Pipelines → Fabric Data Factory (same pipeline engine, new interface)

Sequencing the migration

1. 1Set up a Fabric workspace and configure OneLake - this is the foundation everything else connects to
2. 2Migrate pipelines first - they're the lowest risk and highest frequency of change
3. 3Migrate the SQL warehouse next - export the schema and data, validate query performance, run parallel for two to four weeks
4. 4Migrate Spark workloads last - they typically require the most rework if there are custom library dependencies
5. 5Redirect Power BI reports to Fabric datasets after the warehouse migration is validated

What to watch out for

Dedicated SQL Pool performance differences

Synapse Dedicated SQL Pool uses a massively parallel processing (MPP) architecture that is optimised for different query patterns than the Fabric Warehouse. Run a performance benchmark with representative queries before switching production traffic.

The migration assessment is not optional. We have seen organisations attempt a direct cut-over from Synapse to Fabric without a proper assessment and encounter performance surprises in production.

Timeline and effort

For a typical mid-market Synapse environment, a complete migration takes twelve to sixteen weeks when executed carefully. The Establish phase (assessment and migration plan) takes four weeks. The Build phase takes eight to twelve weeks.

If any of this sounds familiar, talk to us about your data.

Related reading

• The end of the visual hack
• Microsoft Fabric vs Azure Synapse: What UK Mid-Market Teams Need to Know
• Pyramid, Power BI or Fabric: Which One, and When