Please start any new threads on our new site at https://forums.sqlteam.com. We've got lots of great SQL Server experts to answer whatever question you can come up with.

 All Forums
 SQL Server 2000 Forums
 Analysis Services (2000)
 Dimensions and Measures

Author  Topic 

Arnold Fribble
Yak-finder General

1961 Posts

Posted - 2001-11-19 : 07:25:14
OK, here's an OLAP question! Why do most OLAP models partition the columns in the fact table into dimension and measure? Many things, age for example, are equally useful as a dimension (sales of product groups by age-bands) or measure (average age for buyers of these product groups across time). Is this just a side-effect of using MOLAP storage mechanisms, or does it run deeper into dimensional modelling as a whole? Will it change in the future?


Tim
Starting Member

392 Posts

Posted - 2001-12-03 : 01:03:45
"OK, here's an OLAP question! Why do most OLAP models partition the columns in the fact table into dimension and measure? "

Well I don't think there is a good answer except that it is a good practice which:

* minimises the size of your database
* allows fastest processing of cube
* keeps it easy to understand/read/work with
* you will find SQL Server Analysis Services easier to use with strict star or snowflake designs

What you put in the fact table really depends on your requirement.

In your example, if you want *exact* age averages based on age at the time of sales transaction then I would include age as a measure on your fact table.

If you were happy with showing sales by ranges of age at time of transaction, then I would put a key to the age dimension on your fact table, and the actual age value and it's related properties in the Age dimension table.

If you were happy with showing lifetime sales of customers by their current age (and/or ranges) then I would add (or relate) age to your customer dimension and have customer key only on the fact table.

Or some combination of above if several requirements fit.


As for "will it change?" who knows? I do know that the basics of relational design have not changed a whole lot since date & codd. I think the star and snowflake design will prove to be equally as enduring.

hth



-------------------------------
Visit nr's cursor wonderland
Go to Top of Page

Arnold Fribble
Yak-finder General

1961 Posts

Posted - 2001-12-03 : 04:53:56
To explain, this question was posted to fill a gap some weeks ago when the Analysis Services folder had just been created and was devoid of questions. It was asked more as a philosophical than a practical question. I had been reading Pedersen and Jensen's 1998 paper Multidimensional Data Modeling for Complex Data: having done a small amount of clinical data warehousing myself, I could appreciate some of their points.

Reference:
http://citeseer.nj.nec.com/bachpedersen98multidimensional.html


Edited by - Arnold Fribble on 12/03/2001 04:57:44
Go to Top of Page

Tim
Starting Member

392 Posts

Posted - 2001-12-03 : 05:05:28
oh so you're not a newby... sorry 'bout that!

What do you think of my response?

Am i talking rubbish or making sense?


-------------------------------
Visit nr's cursor wonderland
Go to Top of Page

Arnold Fribble
Yak-finder General

1961 Posts

Posted - 2001-12-03 : 06:28:10
As a practical answer to a practical question, it looked fine to me. But as I say, I've only done a small amount of data whore^Wwarehousing.


Edited by - Arnold Fribble on 12/03/2001 06:29:14
Go to Top of Page
   

- Advertisement -