Archives of the TeradataForum

Message Posted: Fri, 17 Aug 2007 @ 14:55:23 GMT

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

We have two columns on every table to indicate when a row has expired and when it has been revised. We use these rows with effective and transaction dates to date chain the temporal aspect of the table. For example, effective/expiration and transaction/revision returns one row for any one entity. In a new data model, we may be doing this on a 2 billion row table (about 1Tb).

We have always updated these rows.

Recently, it was suggested that we could pull the rows that we typically would update into a holding table, update the holding table, delete the same rows from the target table and reinsert the updated rows from the holding table.

Can anyone comment on either/both strategies. We are considering the impact on rollback, load on the box, impact to the user etc....

Thanks for your help.

Jim

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