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!!!

 

[itmonk]

Hi,,

As far as ur first question is concerned, I would suggest a relational mode, even though teradata supports the star and snow flake schemas in dimensional model.

I would need some more time to answer ur second quest.

RK (Teradata Certified Professional)

 

[rrrkrishnan]

Thanks for your response. I am in favour of the relational model as well. Would be glad to hear from you about my second question.

 

[renu123]

For the second part of your question, I think you should go by the second option i.e. 'A type I dimension table and a SCD table holding just the attributes that can change.'
Loading is easy as you only have to updates in the 25 col table. For the other 5 col. table, you can add a row whenever there is a change in an attribute value.

Regards