Atomic Upserts in Teradata

[rwierdsm]

All,

Has anyone had any experience using the Upsert capabilities of the TeraData MultiLoad Stage, using either MultiLoad or TPump?

The TeraData documentation suggests that the transaction is quite efficient, having implemented a feature called Atomic Upserts.

http://www.teradataforum.com/teradata_p ... 061a_6.pdf page 3-196

I am very interested in what the performance might be of these upserts vs. splitting the data into separate insert and update streams with the associated lookups, lookup key maintenance, etc. required.

Responses much appreciated,

Rob