Question

You manage an import-mode semantic model in a Fabric Premium workspace. The model has grown to approximately 12 GB after a full refresh, and users report that XMLA-based partition refreshes fail intermittently and the model occasionally cannot be loaded. You need to allow the model to exceed the default 10 GB in-memory size limit and support incremental partition operations reliably. What should you do first?

Accepted Answer

Enable the Large semantic model storage format setting on the semantic model.

Answer

Convert the model to Direct Lake mode so data is read directly from Delta tables.

Answer

Turn on Automatic aggregations to reduce the in-memory footprint of the model.

Answer

Reduce the refresh frequency and disable the XMLA read/write endpoint for the workspace.