You are managing a set of Dataflow Gen2 queries that are currently ingesting tables into a Fabric lakehouse. You need to ensure that the tables are optimized for Direct Lake connections that will be…

Applying an incremental refresh policy to the semantic model refreshes optimizes the *refresh process* for the semantic model itself, not the underlying lakehouse tables for Direct Lake query performance.

The VACUUM command is used to remove old, unreferenced data files from a Delta table's history, reducing storage size, but it does not directly optimize for Direct Lake read performance like V-Order does.

Direct Lake mode in Fabric semantic models achieves high performance by directly reading data from Delta tables in the Lakehouse. Applying V-Order through the OPTIMIZE command significantly enhances Direct Lake query performance by organizing the Delta table data files in a specific way that aligns with the needs of analytical workloads, improving scan efficiency and reducing I/O.

Using shortcuts allows semantic models to reference lakehouse tables, but it's a method of connection, not an optimization technique for the physical data storage that enhances Direct Lake performance.