Use SQL Server Management Studio to Connect to DataBricks?

[Shawn_Eary]

The Notebook UI doesn't always provide the best experience for running exploratory SQL queries. Is there a way for me to use SQL Server Management Studio (SSMS) to connect to DataBricks? 

See Also:
https://learn.microsoft.com/en-us/answers/questions/740986/how-to-connect-from-ssms-or-azure-data-st...

1. I know I can use Visual Studio Code and Power BI, but I want to be able to use SSMS if that's possible.
2. There are a few lesser-known SQL Monitor tools that use JDBC to connect to DataBricks, but for various reasons some organizations may hesitate to approve employee use of such tools

[-werners- · Accepted solution]

What you can do is define a SQL endpoint as a linked server.  Like that you can use SSMS and T-SQL.
However, it has some drawbacks (no/bad query pushdown, no caching).
Here is an excellent blog of Kyle Hale of databricks:

Tutorial: Create a Databricks SQL Linked Server in SQL Server | by Kyle Hale | Medium

Tutorial: Create a Databricks SQL External Data Source in SQL Server with PolyBase | by Kyle Hale | ...

If you are in the MS ecosystem, his blog is definitely worth checking out.

[Shawn_Eary]

That looks like a pretty good solution, but some organizations aren't going to give developers the permission to implement something like that.

Also, not only is it a pain to connect SQL Server Management Studio (SSMS) to Databricks, but right now, it seems to also be a pain to connect Microsoft Power Apps to Databricks. I'm not a fan of Microsoft Power Apps personally, but some organizations practically mandate use of Microsoft Power Apps.

Microsoft doesn't seem to be particularly willing to help with these incompatibility issues. I think they would rather push Fabric instead of promoting Azure Databricks.

If Databricks, Inc. isn't interested in improving compatibility with Microsoft tools, they may wind up losing customers. 

[-werners-]

I think MS wants to make a closed ecosystem for data.  If Databricks wants to go all-in on that it would mean dropping AWS/GCP. Bad decision IMO.