Archives of the TeradataForum

Message Posted: Mon, 28 Mar 2005 @ 18:15:09 GMT

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

Hi,

we purge about 30.000.000 rows ( one week data ) once a week out of a 15 month transaction table ( about 2 Billion rows )( CDRs ).

We use Multiload delete and the where clause is "where start_date <= date'YYYY-MM-DD'. start_date IS NOT the PI. This runs only about 30 minutes on a "4 node coexistancy machine 46 AMPS".

But make sure, that there are no SI's defined on the table while the Mload delete is going on. We found, that dropping the SI's and recreating them after the mload delete makes a better overall performance.

Greetings

Christian

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