if SQL executed on one node, what is the algorithm for selecting the execution node?
If all the tables queried by SQL are replicated tables, is the SQL
executed on all nodes or only on one node?
