data modeling question

[rrrkrishnan]

I have a couple of questions.

1. Which is best suited for a datamart that will exist as a teradata database. The dimensional model or relational model.

2. We have an account table (2.5 million records) which has around 30 cols of which say 25 are stable (or we do not care for history on these cols) and 5 cols which can change and the changes have to be tracked. The load is done once a month and we also have a customer metrics table which holds monthly data and has info. like current balance amt, avg
balance amt etc.

Which would be the best way to implement this.

1. A SCD table
2. A type I dimension table and a SCD table holding just the attributes that can change.
3. multiple history tables (one for each attribute).
4. Other (Please explain.)

Any response is much appreciated.

Thanks!!!

 

[farhy]

Hi,

1) that is mostly a question of favor, Mr Kimball he would recommend the denormalized schema,
NCR recommends an ER Modell, with Join Indexes, Global Temp. tables, etc. (even for datamarts )

From my point of view the denormalized design is suitable for datamarts.

2) Read the Design Manual from NCR, there is technique
calles vertical partitioning:
Use 2 Tables: one with the stable 25 columns and one with
5 columns, and use the same Primary index in both
tables, or a join index.

hope that helps.

so long
farhy