Archives of the TeradataForum

Message Posted: Thu, 28 Feb 2008 @ 14:01:42 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

MLOAD does its "uniqueness" checking at the beginning of Phase 4. It sorts the data on each AMP and gets rid of duplicate (based on sequence value - unlike Fastload that has to compare). Then and only then does it apply the new data to the table. If there are any duplicate UPI values, anything after the first one goes into the error table - no duplicate row check at the table.

What we do not know about this exercise is whether or not there are rows in the second error table (UV). Any insert into the error table(s) is a row at a time insert and of course this will be slower. If the IGNORE error treatment is used on the .DML then this is not a factor either.

However, making it a multiset table with a NUPI is not a good knee jerk reaction.

Just my thoughts on the topic.

Michael Larkins
Certified Teradata Master
Certified Teradata SQL Instructor

	<Prev	Next>		<<First	<Prev	Next>	Last>>