Rows will be distributed among the AMPs based on PI value.
But if i consider that the table is fall back protected and I insert a row in it, then the row will go to appropriate AMP and get stored(say in AMP-0). but since the table is Fall Back protected, copy of primary row will be maintained in Fall back AMP. If i consider that there is 4 AMP cluster then where the Fall back row will be stored i.e. whether in AMP-1, AMP-2 or AMP-3 ? How the Fall Back Amp will be decided, whether any Hashing mechanism is used in it??
it's using the same hashing algorithm, but a different hashmap, the Fallback Hashmap.
Check the Database Design manual:
Chapter 8: Teradata Database Indexes and Partitioning
I searched this Manual in "http://www.teradatamagazine.com/" after registering to it but I cannot find such manual to refer. Where actually i can get such Teradata manuals? Please help me out.