Question

You manage a Direct Lake semantic model on an F64 capacity. Users report that certain visuals are intermittently slow during peak hours, but the same visuals are fast at other times. Direct Lake fallback to DirectQuery is NOT occurring. You suspect memory pressure is causing columns to be paged out. Which behavior of Direct Lake best explains the intermittent slowness, and what should you investigate first?

Accepted Answer

Direct Lake loads (transcodes) column data into memory on demand and evicts least-recently-used columns under memory pressure; investigate the model's column footprint and capacity memory limits to reduce cold-load reloading.

Answer

Direct Lake caches the entire Delta table in memory permanently after the first query; investigate whether the scheduled refresh is disabled during peak hours.

Answer

Direct Lake always queries the SQL analytics endpoint at runtime; investigate the endpoint's compute autoscale settings to add more replicas.

Answer

Direct Lake performance depends only on the number of active relationships; investigate whether bidirectional relationships are causing ambiguous filter paths.