Microsoft Fabric Data Engineer Associate DP-700 Practice Question

ROUND_ROBIN distributes rows without regard to key values, so joins with other tables typically require large data movement. Redistributing the large fact table by hashing on the join key (CustomerID) colocates matching rows on the same distribution, allowing the join to be performed locally and greatly reducing shuffle cost. REPLICATE would eliminate data movement but is intended for small tables and would incur excessive storage for a 5-billion-row fact table. Creating a clustered index does not address distribution-based data movement, and result-set caching benefits repeat executions only when the underlying data is unchanged; it does not eliminate movement in the execution plan itself.