Archives of the TeradataForum
Message Posted: Tue, 09 Jan 2001 @ 06:46:00 GMT
Thanks for the clarification John. However, if you ever saw detailed transactions from retail with transaction types etc. It's probably not that different.
Maybe we should look at this in two ways? First is the 'model' - i.e., the 'logical' modelling of data - vs. - the physical layout of the data. As most of us are familiar with logically modelling data and the one-to-one (almost) physical layout, are we not forgetting the value of Teradata with UPI/NUPI index assignment? If I have a 'main event table' with a selected UPI (maybe the PK), it would behoove us to assign the same FK columns in the subtypes as NUPI's, any joins between the main event and any sub types would be extremely efficient. I have used this in many instances. HOWEVER, there is value in identifying major attributes that you could 'roll up' to the main event to reduce extremely efficient joins. Creating effective views over this scenario would make it easier from the user point of view - if that was a concern.
What ever, it's always a trade off.
Just my $.02 or tu-pence worth!
|Copyright 2016 - All Rights Reserved|
|Last Modified: 28 Jun 2020|