@Bogdan_Teleuca  The following statements are based on Viya 4 but I believe it's the same for Viya 3.5.   "Tables are loaded, then replicated, copied on all workers (DUPLICATE=YES option in the data step). Again, this is OPTIONAL, but it seems to improve query performance." Afaik that's because all data blocks exists on all nodes and therefore no data movement between nodes is required.   "The rest of the tables are loaded normally. Copies=0 is used to reduce redundancy and hence, data movement." I believe that's not true. Less copies means lower memory consumption but potentially more data movement with a cost to performance. Because the dimension tables are replicated such movement won't be necessary for the join in the view as such but depending on aggregations using the view less copies could lead to more data movement (like aggregation by a variable by which the underlying fact table is not partitioned). ... View more