I am just wandering, in which part of logical architecture of teradata does hashing happens. I read somewhere when data loading happens first data is transferred in AMPs and then hashing happens and rows are sent to respective AMPs according to their hash values. My question is does hashing happens at PE or at AMP ?
Could be either AMP or PE. In particular, FastLoad sends data blocks to AMPs round-robin and the AMPs hash the rows and send to the correct target AMP. But for a single-row fetch or insert, the PE computes the hash to determine which AMP should handle the request.
What I know, hash algorithm present at the PE level. And hash maps are present in Message Parsing Layer using which data get inserted into the respective AMP.