Archives of the TeradataForum

Message Posted: Fri, 05 Jan 2007 @ 18:30:11 GMT

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

I just finished doing an EDW implementation utilizing DataStage Enterprise Edition against Teradata and I can tell you that there is nothing inherent to DataStage that would require you to place a UPI on all of your tables.

As others that have responded to your initial posting have commented on, implementing a UPI on every table will give you serious performance issues when it comes to utilizing the data.

I would go back to your DataStage development team and find out their reasoning for insisting on a UPI and see if you can work through any issue with them to find out why they think they need a UPI. Remember that an EDW needs to be a balance between the needs of ETL and the needs of the applications that will be utilizing the data. You can't let either one dictate your design.

A couple of things I discovered about DataStage and Teradata from that last project:

- Use the Teradata API stage (which uses CLI) instead of ODBC stage if you can't use one of the load utilities (MultiLoad, FastLoad, TPump). The Teradata API stage is much quicker at everything then the ODBC stage.

- If you are going to be doing a large volume of Upserts then hand code them as Atomic Upserts instead of letting the stage do it for you. If the stage does it they are coded as updates and inserts instead of Atomic Upserts.

Robert Meunier

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