 |
 |
Archives of the TeradataForum
Message Posted: Mon, 08 Jan 2001 @ 20:33:40 GMT
| Subj: | | Re: Technical Data Modelling |
| |
| From: | | Frank Martinez |
Hello!
I came off a project in Mexico just last year where the client had decided that a denormalized design (based on an original DB2
warehouse) was the basis of the Teradata design. And IT STANK! We kept trying to go back to a decent NORMALIZED design TO SOLVE THEIR
PERFORMANCE PROBLEMS! There turned out to be a real need, also, to do the statistical analysis for TWO or THREE large tables (out of
several hundred, by the way), in order to select the correct indices. I could see where doing the Data Access model would also allow you to
preplan using join indices, once you know that important queries take too long and can't be solved some other way. But the basis of a good
Teradata design is a normalized model based on a logical model that addresses the client's business questions.
iv
Frank C. Martinez IV
Groupe Bull - Integris
Frank C. Martinez IV
| |
