Archives of the TeradataForum

Message Posted: Tue, 01 Mar 2011 @ 10:58:40 GMT

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

Thanks guys for your inputs.

Dieter, please correct me if I'm wrong: in my vision, the delete was a fastpath. I have few reasons to think so:

1. "We do an all-AMPs DELETE of 518 partitions" instead of something like "We do an all-AMPs DELETE from...". I might be wrong here though.

2. It was the only statement in a transaction, there is no JI nor PK on the table.

3. The delay was in End Transaction Step, not the DELETE step. Actually, that's what hit me. What can take the Edt step so long and why does it need so many IOs and CPU? Can it be due to secondary indexes positioned on the table? AFAIK, handling indexes when executing DML statements is done in the INSERT/UPDATE/DELETE steps (together with table data), not in the END TRANSACTION step. Which seems logical to me too.

Tomas

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