Azure Synapse Analytics
Azure Synapse Analytics is Azure's earlier analytics service, combining dedicated and serverless SQL pools, Apache Spark pools and Data Factory-style pipelines in one workspace. It is still supported, but Microsoft's guidance is to prefer Microsoft Fabric for new investments, and it provides migration tools.12
- Intended for
- Teams running existing Synapse workloads, Data engineers, Data warehouse developers and DBAs, Data platform architects planning migrations
- Runs on
- Azure regions as a PaaS workspace in your subscription
At a glance
- What is it?
- Azure Synapse Analytics is an enterprise analytics service that brings data warehousing (Synapse SQL), big data processing (Apache Spark pools), data integration pipelines and links to Power BI, Azure Cosmos DB and Azure Machine Learning into one workspace, managed through Synapse Studio. Synapse SQL offers dedicated SQL pools, which reserve compute for data in SQL tables, and a serverless SQL endpoint for ad hoc queries over files in the data lake. Its overview page (Learn date 10 Jul 2024) still lists Data Explorer as a preview component, but Microsoft set that component's retirement for 7 Oct 2025.13
- What does it do?
- Teams load and model warehouse data in dedicated SQL pools, query Parquet, CSV, TSV and JSON files in the data lake with serverless SQL or Spark, and build ETL and ELT pipelines with the same integration engine as Azure Data Factory (90+ sources, code-free data flows). Spark pools support SparkML, Delta Lake, .NET for Spark and GPU-accelerated nodes. Synapse Link for Azure Cosmos DB gives near real-time analytics over operational data without ETL.14
- Who needs it?
- Organizations that already run Synapse dedicated SQL pools, Synapse Spark or Synapse pipelines in production and must keep them running while they plan their next step. Microsoft also lists cases where Synapse Spark still fits better than Fabric Spark: GPU-accelerated pools, an external Hive Metastore, JDBC connections and .NET for Spark (C#).4
- What does it need?5
- An Azure subscription with a Synapse workspace
- An Azure Data Lake Storage Gen2 account as primary storage
- For migration planning: an inventory of dedicated SQL pools, Spark pools, notebooks, Spark job definitions and pipelines
- For moves to Fabric: a Fabric capacity and Microsoft Entra identities to replace SQL authentication users
- What it is not
- Synapse is not where Microsoft wants new analytics projects to start: its Azure Architecture Center says to prefer Microsoft Fabric for new investments. Synapse is also not the same product as Fabric Data Warehouse or Fabric Spark, even though Microsoft documents migration paths between them. Synapse Data Explorer is no longer an option; Microsoft set its retirement for 7 Oct 2025, said workloads would be deleted, and recommends Eventhouse in Fabric instead.23
Availability and licensing. Available and supported; the Synapse overview page carries no retirement notice for the service as a whole (checked 26 Oct 2026). Microsoft's Azure Architecture Center (Learn date 30 Jul 2026) says to prefer Microsoft Fabric for new investments. Synapse Data Explorer (preview) was set to retire on 7 Oct 2025, with its workloads deleted. Microsoft's Cloud Migration Factory program offers selected customers hands-on migration help to Fabric at no cost for 6 to 8 weeks.1
The problem it solves
Before Fabric, an Azure analytics estate needed a warehouse, a Spark engine and an integration tool. Synapse combined them in one workspace with shared security and one studio, over data in Azure Data Lake Storage.
For current Synapse customers the main question is how long to stay and how to move. Microsoft documents where Fabric and Synapse differ, offers a Migration Assistant for dedicated SQL pools and guides for Spark pools, notebooks and metadata, and runs a Cloud Migration Factory program that gives selected customers hands-on help moving to Fabric.34
How it works
- Synapse SQL: a distributed T-SQL engine with dedicated SQL pools (reserved compute) and a serverless endpoint over lake files; PREDICT scores machine learning models in T-SQL.
- Apache Spark pools: a pool is a fixed compute boundary of 3 to 200 nodes from which each job provisions a cluster; memory-optimized or GPU-accelerated node families.
- Pipelines: the Azure Data Factory integration engine inside the workspace.
- Storage and metadata: Azure Data Lake Storage Gen2 as primary storage, with an internal Hive Metastore or an external one in Azure SQL Database.
- Synapse Studio: one interface to ingest, explore, orchestrate, monitor and manage access across SQL and Spark.4
Diagram as a list
Copilot & business apps
- Synapse StudioOne interface to ingest, explore, orchestrate and monitor
- Power BIReporting over Synapse SQL
Data & analytics
- Dedicated SQL poolsReserved T-SQL warehouse computeConnects to Azure Data Lake Storage Gen2
- Serverless SQL endpointOn-demand T-SQL over files in the lakeConnects to Azure Data Lake Storage Gen2
- Apache Spark poolsData engineering and machine learning on Spark
- Pipelines (Data Factory engine)ETL and ELT orchestration from 90+ sources
- Azure Data Lake Storage Gen2Primary storage for the workspace
- Hive Metastore (internal or external)Table metadata for Spark
Capabilities
Dedicated and serverless SQL pools15
Dedicated pools reserve processing power for predictable warehouse workloads; the serverless endpoint queries lake files on demand for bursty or ad hoc work.
Why it matters: Lets one workspace cover both a structured warehouse and exploration of raw files.
Limits: For new warehouses Microsoft prefers Fabric; the Fabric Migration Assistant converts dedicated pool schemas, but T-SQL compatibility is not complete.
Apache Spark pools14
Spark pools with SparkML, Azure Machine Learning integration, Delta Lake support, .NET for Spark and GPU-accelerated node sizes.
Why it matters: Still the Microsoft option for GPU Spark pools, external Hive Metastore and C# Spark code.
Limits: Synapse Spark runtimes listed by Microsoft go up to Spark 3.4, while Fabric offers newer runtimes; Synapse has no high-concurrency sessions or V-order.
Integrated pipelines1
The same data integration engine and experience as Azure Data Factory, with 90+ sources, code-free data flows and orchestration of notebooks, Spark jobs, stored procedures and SQL scripts.
Why it matters: ETL and ELT can be built without leaving the workspace.
Limits: Pipelines move to Fabric Data Factory only through migration work.
Synapse Link for Azure Cosmos DB6
A hybrid transactional and analytical capability that analyzes Cosmos DB operational data in near real time without ETL.
Why it matters: Avoids building pipelines to analyze operational NoSQL data.
Limits: Cosmos DB now also offers Fabric mirroring for the same no-ETL pattern.
Documented migration path to Fabric4
Microsoft publishes a Spark feature comparison, migration guides for pools, configurations, libraries, notebooks and metadata, and a Migration Assistant for dedicated SQL pools.
Why it matters: Teams can plan a phased move rather than a rewrite.
Limits: Fabric lacks GPU pools, external Hive Metastore and .NET for Spark, so some workloads need rework first.
Practical use cases
A dedicated SQL pool runs core reporting and the team must decide its future.
- Approach
- Inventory the objects, then upload a DACPAC to the Fabric Migration Assistant to see which objects convert and which need fixes.
- Role of Azure Synapse Analytics
- Synapse remains the running system while Fabric is tested as the target.
- Data, infrastructure and skills
- A DACPAC export and a Fabric trial or capacity.
- Type of benefit
- Lower migration risk
- Caveats
- T-SQL parity is incomplete; external tables, multi-statement table-valued functions and SQL authentication need changes.
- First step
- Export a DACPAC of the dedicated pool and run it through the Migration Assistant.
Spark jobs depend on GPU pools or C# code.
- Approach
- Keep those jobs on Synapse Spark for now and rewrite .NET for Spark code in Python or Scala before any move.
- Role of Azure Synapse Analytics
- Synapse Spark continues to run the workloads Fabric cannot host yet.
- Data, infrastructure and skills
- An inventory of GPU and .NET dependencies.
- Type of benefit
- Business continuity
- Caveats
- Running two platforms adds cost and operational overhead.
- First step
- Tag every Spark job that uses GPU pools, external Hive Metastore or .NET for Spark.
Operational data in Azure Cosmos DB needs near real-time analytics.
- Approach
- Use Synapse Link for existing setups; for new work, evaluate Fabric mirroring of Cosmos DB.
- Role of Azure Synapse Analytics
- Synapse queries the Cosmos DB analytical store without ETL.
- Data, infrastructure and skills
- A Cosmos DB account with analytical store enabled and a Synapse workspace.
- Type of benefit
- Fresher analytics
- Caveats
- New investments are better placed in Fabric per Microsoft's guidance.
- First step
- Compare one Synapse Link query with the same data mirrored into Fabric.
Works with
Optional integration
- Power BISynapse lists deep integration with Power BI.
- Azure Cosmos DBSynapse Link gives near real-time analytics over Cosmos DB data.
- Azure Machine LearningSynapse Spark integrates with Azure Machine Learning.
Alternative approaches
- Microsoft FabricMicrosoft says to prefer Fabric over Synapse for new investments.
Optional integration for
- Azure Cosmos DBAzure Synapse Link gives near real-time analytics over Cosmos DB data.
- Power BIAzure Synapse lists deep integration with Power BI.
- Azure Machine LearningAttached Synapse Spark pools provide Spark processing.
Relationship labels follow Microsoft's documentation. "Alternative approaches" does not mean one is better: each profile says when it fits.
Compared with
Getting started
Inventory the estate
List dedicated SQL pools, serverless queries, Spark pools, notebooks, Spark job definitions, pipelines and linked services.
Check: Every workload has an owner and a keep, migrate or retire decision.
Check Spark feature gaps
Compare each Spark workload against Microsoft's Fabric and Synapse Spark comparison (GPU pools, external Hive Metastore, .NET for Spark, runtime versions).
Check: Workloads that cannot move yet are flagged.
Pilot the warehouse move
Run the Fabric Migration Assistant on a DACPAC or a direct connection and fix the objects it flags, starting with primary objects.
Check: The migrated schema deploys to a Fabric warehouse without errors.
Copy data and test
Copy data with a Fabric Data Factory copy job, then compare results between the old and new warehouse.
Check: Row counts and key reports match.
Reroute and decommission
Point applications and reports at the new warehouse, then retire the Synapse objects.
Check: No queries reach the old pool for a full business cycle.
Official resources
Deciding whether you need it?
Three free tools apply Microsoft's own published guidance to your situation, and say when Azure Synapse Analytics or any new technology is not needed.
Sources
Each statement above links to the source it comes from. Labels say who reported it.
- What is Azure Synapse Analytics? (opens in a new tab)
- AI and machine learning products (Azure Architecture Center) (opens in a new tab)
- What is Azure Synapse Data Explorer (Preview) (opens in a new tab)
- Compare Fabric and Azure Synapse Spark (opens in a new tab)
- Migration Assistant for Fabric Data Warehouse (opens in a new tab)
- Azure Cosmos DB overview (Unified AI Database) (opens in a new tab)
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.