 |
 |
Archives of the TeradataForum
Message Posted: Sat, 12 Apr 2009 @ 00:47:25 GMT
| Subj: | | Re: Identity function sequence number |
| |
| From: | | Fred Newbrough |
I agree with Victor. We used surrogates in our ETL design but decided against using identity columns for the reasons he mentioned and instead
developed a procedure to create them based on specified source table columns. These keys are stored in an autogenerated key table or in a
prexisting key table. This approached offerred us the greatest degree of flexibility and the benefit of meaningful keys that improve query
performance rather than identity columns which can hinder performance while providing the illusion of good distribution which is actually
distribution that is meaningless to the actual internal distribution of the data.
| |
