Azure Synapse vs Azure Data Factory: A Comparison

"Should we use Synapse or Data Factory?" is one of the most common questions about Azure's data services, and it is slightly the wrong question. It treats as rivals two services that do different jobs. One moves data. The other stores and analyses it. The confusion exists because the second one contains a copy of the first one's engine. Azure Synapse vs Azure Data Factory is less a contest than a question of scope.

Two definitions

Azure Data Factory is, in Microsoft's words, a managed cloud service built for hybrid extract-transform-load, extract-load-transform and data integration projects. It creates data-driven workflows that orchestrate data movement and transform data at scale (Microsoft Learn: introduction to Azure Data Factory).

Azure Synapse Analytics is an enterprise analytics service that brings together SQL technologies for data warehousing, Spark for big data, and pipelines for data integration (Microsoft Learn: what is Azure Synapse Analytics).

Read the second definition again. Pipelines are one item in the list. Synapse includes data integration; it is not mainly a data integration service.

What each one can do

The first three rows are the overlap. Below them, the services part ways.

Data Factory holds no data and has no query engine. This surprises beginners. It connects, copies, and tells other systems to do work: run this stored procedure, run that notebook. Its own compute, the integration runtime, moves data and runs data flows. When a pipeline finishes, nothing is "in" Data Factory.

Synapse is where data lives and is queried. A dedicated SQL pool holds tables. A serverless pool reads the lake. Spark pools run code. The pipelines inside it are how data gets in and how those engines are sequenced.

The overlap: pipelines

The pipelines in a Synapse workspace use the same engine as Data Factory, a fact Microsoft states in both products' documentation. Pipelines, activities, triggers, connectors and data flows work the same way.

They are not identical. A short list of features exists on one side only: Data Factory has the SSIS runtime, sharing of self-hosted runtimes, global parameters and the Power Query activity; Synapse has monitoring of the Spark jobs behind data flows (Microsoft Learn: differences from Azure Data Factory). Synapse pipelines vs standalone ADF goes through each.

So the narrow version of the question, where should our pipelines live, has a real answer that depends on those features. The broad version does not, because Data Factory cannot replace the rest of Synapse.

How each is sized and paid for

The cost models differ as much as the purposes.

Data Factory is consumption-based throughout. Microsoft describes it as serverless, with no fixed-size compute to plan for. The meters are activity runs for orchestration, Data Integration Unit hours for copies, and vCore hours for data flows (Microsoft Learn: plan to manage costs for Data Factory). A factory with no runs costs almost nothing.

Synapse has a different model for each engine. Its two SQL engines show the range:

• Dedicated SQL pool: compute you provision, measured in data warehouse units, billed while it runs. It can be paused, leaving only storage to pay for (Microsoft Learn: dedicated SQL pool architecture).
• Serverless SQL pool: no reserved resources; you are charged for the data your queries process (Microsoft Learn: serverless SQL pool).

The practical consequence: a forgotten data factory is cheap, and a forgotten dedicated SQL pool is not. Anyone learning on a personal subscription should pause or delete dedicated pools at the end of each session.

How they fit together

In a typical Azure analytics platform the two are stages, not alternatives:

1. Ingest. Pipelines copy data from source systems, databases, files, APIs, into the data lake.
2. Store. Raw and refined data sits in Azure Data Lake Storage.
3. Transform. Spark, SQL or data flows clean and model it.
4. Serve. A warehouse, such as a dedicated SQL pool, holds the modelled tables that reports query.
5. Orchestrate. Pipelines run the above in order, on a schedule.

Data Factory can do steps 1 and 5, and step 3 with data flows. Synapse can do all five. That is the whole relationship. ETL vs ELT explains the pattern behind steps 3 and 4.

Microsoft's own Data Factory introduction describes exactly this arrangement: a company ingests logs and reference data, processes them on a Spark cluster, and publishes the result into a cloud data warehouse such as Azure Synapse Analytics for reporting, with Data Factory automating the workflow.

Questions that decide it

Do you need a place to store and query data? If yes, Data Factory alone is not enough. You need a warehouse or lake engine: Synapse, or Azure SQL Database, Databricks, or another.

Is your only need to move data between systems on a schedule? Data Factory alone is the lighter choice. There is no reason to create a Synapse workspace to copy files.

Is everything going to live in one Synapse workspace? Then Synapse pipelines keep it in one place, with one deployment and one access model.

Does integration serve many destinations, of which Synapse is only one? A standalone data factory is the more natural centre.

Do you have SSIS packages to keep running? Data Factory is where Microsoft documents running them without caveats.

The third option

No comparison of these two is complete without the service that is replacing both in Microsoft's guidance. Data Factory's documentation describes Data Factory in Microsoft Fabric as the next generation of Azure Data Factory, and Synapse's SQL pool documentation advises those new to data warehousing to start with Fabric Data Warehouse. Microsoft Fabric vs Azure Data Factory covers what changes.

The distinction this article draws survives the move. Fabric also separates data integration from the engines that store and analyse data. Understanding which job is which is the part worth learning.

Misconceptions

• "Synapse replaces Data Factory." Only its pipelines overlap, and with gaps.
• "Data Factory is a data warehouse." It stores nothing.
• "Synapse is just a database." It is several engines in one workspace.
• "You must pick one." Many platforms use Data Factory for enterprise ingestion and Synapse for analytics.
• "Synapse pipelines are a cut-down Data Factory." They are the same engine with a handful of differences in each direction.

For each product on its own, see Azure Synapse Analytics explained and the Azure Data Factory roadmap.

Azure Data Factory, Azure Synapse and Azure Databricks are covered in the Data Engineering course.

Sources

• Introduction to Azure Data Factory (Microsoft Learn) — Microsoft Learn
• What is Azure Synapse Analytics? (Microsoft Learn) — Microsoft Learn
• Dedicated SQL pool architecture (Microsoft Learn) — Microsoft Learn
• Serverless SQL pool in Azure Synapse Analytics (Microsoft Learn) — Microsoft Learn
• Plan to manage costs for Azure Data Factory (Microsoft Learn) — Microsoft Learn
• Differences from Azure Data Factory (Microsoft Learn) — Microsoft Learn