Direct Lake over Mirrored Database

[markv_greenleaf]

Can we have a Direct Lake mode semantic model against a Mirrorred Database in Fabric? The Direct Lake docs indicate that it uses a SQL Analytical endpoint as a source. And the Mirroring docs indicate a SQL Analytical endpoint is created on a mirrored database. It would seem reasonable that a Direct Lake model could be created on a Mirrored Database, but an explicit statement would be helpful.

[v-karpurapud]

Hi @markv_greenleaf 

Thank you for reaching out to the Microsoft Fabric Community Forum.
 

As correctly pointed out by @lbendlin  that Direct Lake mode prefers reading from Delta Lake files directly, and only utilizes the SQL Analytics Endpoint as a fallback mechanism.

While the SQL Analytics Endpoint is part of the overall setup, Direct Lake mode does not primarily query from it. Instead, it reads directly from the Delta Lake-formatted files stored in OneLake, which allows for high-performance, near real-time analytics.

The SQL Analytics Endpoint is only engaged when a query becomes too complex or uses unsupported DAX features, at which point the engine gracefully falls back to DirectQuery mode.

For more detailed guidance, please refer to the official Microsoft documentation:

Direct Lake overview - Microsoft Fabric | Microsoft Learn

If you find this post helpful, please mark it as an "Accept as Solution" and give a KUDOS.

Thank You!

[markv_greenleaf]

Thank you, @lbendlin and @v-karpurapud for the replies. Perhaps I should have phrased my question better. 🙂
When the queries need to fall back to using the SQL Analytics endpoint, is the one created with a Mirrored Database supported for that fall back?

Thanks.

[v-karpurapud · Accepted solution]

Hi @markv_greenleaf 

No, the fallback to DirectQuery in Direct Lake mode only works when the semantic model is built over a Lakehouse.

While mirrored databases do expose a SQL Analytics Endpoint, they are not supported as sources for Direct Lake models, and therefore cannot be used for Direct Lake fallback.

Only the SQL Analytics Endpoint of a Lakehouse is used when Direct Lake falls back to DirectQuery.

Thank You!

[markv_greenleaf]

Thank you, @v-karpurapud That is what I was looking for. 

[lbendlin]

The Direct Lake docs indicate that it uses a SQL Analytical endpoint as a source

Not primarily. The preferred option is to go to the Delta Lake files directly, only if your query becomes too complex would it fall back to Direct Query mode.