Archives of the TeradataForum
Message Posted: Tue, 27 Feb 2001 @ 13:08:49 GMT
In our specific case:
We convert three 3rd normal fact tables (all with different Primary Indexes) (~200 million record, ~60 million records, ~5 million records respectively) into separate (by geography) star schema databases. Each database has an atomic table (with a new common 6 part key UPI) and 8 summary tables (built from the atomic table.) The dimension tables for each star schema are made up of data from ~50 3rd normal form supporting tables. We provide the user with ~1.1 million ways (tuples) of looking at this data with an average response time of ~15 seconds.
This design has been "blessed" by NCR as a necessary evil to support the strict response times imposed on us by the users of our application given the amount of disk space and processing power we currently have available to make "lemonade" with.
Sometimes de-normalization is necessary as a transitional solution to a larger problem.
|Copyright 2016 - All Rights Reserved|
|Last Modified: 28 Jun 2020|