Archives of the TeradataForum
Message Posted: Fri, 02 Nov 2001 @ 13:04:04 GMT
| Subj: | | Re: Data Model with Microstrategy as frontend tool |
| |
| From: | | Cornelius Meade |
Interesting conversation....
My experience has been that while these tools do not always generate the most efficient SQL code/queries in all instances, 3NF is not
significantly different than a starish model on a pure performance basis all else being equal (which is difficult to achieve!). I will say
however that Teradata's architecture can allow a given schema to perform where you might have more trouble getting equal performance out of
that same schema running on a non-Teradata platform.
In my view, it is the understanding/knowledge that the end users have about the business data and its structure that makes the
determining factor. Stars, snowflakes and schemas of their like can be complex or simple. A 3NF schema can be complex or simple.
Generally (and this is a big generalization) I find that star schemas tend to be a little less complex for business users/non-modelers to
grasp as the business rules for relating data items to one another is a little more intuitive/clearly shown. The OLAP tool tie
in/generalization is that this tends to be a more straight forward environment to operate in for the tool and well as the user.
A 3NF schema for the same business data generally will be a little more complex (most likely expressed in the terms of the total number
of tables and more relationships) but might be a little more flexible in terms of the ways users can breakdown the data and to evolving
business needs. Queries in 3NF schemas tend to be a little more complex on average because of the need to join more tables (this is a
generalization also). This translates into a little more complexity as a rule, that the users/tools have to be able to handle.
My musings....for what its worth.....Neil
| 
