Question

You are building a Direct Lake semantic model over a Fabric lakehouse. The lakehouse contains several Delta tables written directly by a Spark notebook. Your team wants the semantic model to read table data with the least dependency on additional metadata-syncing layers, and you want to avoid the automatic framing delays that can occur when the SQL analytics endpoint has not yet caught up with newly written Delta files. Which approach should you choose when creating the model?

Accepted Answer

Create a Direct Lake on OneLake model that connects directly to the lakehouse Delta tables in OneLake

Answer

Create a Direct Lake on SQL analytics endpoint model built from the lakehouse SQL endpoint

Answer

Create an Import mode model and schedule refresh after each notebook run

Answer

Create a DirectQuery model against the SQL analytics endpoint