I am getting kind a confused here.
Ex . USI and NUSI - All it does is -
a. creats a subtable (Q1. not sure where ? if in all amps then the same table is replicated in all amps?)
b. While fectching first, hashmap look up coresponding to the value of USI.( Q. If incase of NUSI haspmap look up is not done then how system identify where (in which AMP) the specisfic column recides?)
a. a SI subtable is distributed just like data table to all amps. and it is not the same table, it is distributed based on row hash.
b. In case of NUSI, the PE sends a broadcast message to all the amps, The required rowid is looked up on NUSI subtable on all the amps and once the amps containing required data rows is identified then only those amps are further processed.