Migrating data from synapse to databricks

[Krizofe]

Hello team,

I have a requirement of moving all the table from Azure Synapse (dedicated sql pool) to databricks.

we have a data coming up from source to azure data lake frequently.

we have Azure data factory to load data (data flow does the basic transformation ) to Synapse.

we are looking to migrate data(tables) from synapse to databricks as a requirement.

And the new data from azure data lake to be ingested directly to databricks bypassing synapse.

can you please let me know the best approach to perform this action.

your insights are valuable!! Thanks

[Rishabh-Pandey]

Use Azure Data Factory (ADF):

• Configure a pipeline in ADF to copy data from Synapse SQL to Azure Data Lake. Set up an ADF Copy Activity to handle this data transfer.

• Source Dataset: Azure Synapse SQL table.

• Sink Dataset: Azure Data Lake Storage or Azure Blob Storage.

• Data Movement Activity: Use the Copy Data activity in ADF.

Load Data into Databricks

1. Create Delta Tables in Databricks:

   • Use Delta Lake for storing and managing data in Databricks. Delta Lake provides ACID transactions and supports efficient queries.

2. Automate Data Ingestion:

   • Set up Databricks jobs or Delta Live Tables to automate the ingestion of data from Azure Data Lake into Delta tables.

3. Update Data Pipeline

   1. Adjust ADF Pipelines:

      • Update your Azure Data Factory pipelines to load new data directly into Databricks instead of Synapse.

      • Source Dataset: Azure Data Lake Storage.

      • Sink Dataset: Databricks Delta table.

[Krizofe]

Hello Rishabh Pandey,

Thanks for the reply but just have some concerns.

The data volume is huge like in billions and data gets deleted once loaded to Synapse.

Pushing all those data back to Azure Storage account from Synapse will be time consuming and costing more on storage. (Note: we have 3k tables with billions of rows)

So, we have any approach to directly move data to Databricks without staging it to Azure storage.

Thanks for the reply, mate.

Your insights will be valuable.

[Rishabh-Pandey · Accepted solution]

@Krizofe  You can use Azure Databricks to directly query and load data from Azure Synapse using Apache Spark. Databricks has built-in connectors that allow it to read from Synapse (or its underlying SQL Data Warehouse).

load it from synapse and then write it to ADLS .