What woudl be the better design (migth be double post)

H

Harry Leboeuf

Hello

In our dimensions we have a lot of data that is related on several levels.

Ex

Table 1 Company_ShowType (PKey Comp_Id, CShowType_Id) Data
(Local_description, ShowType_Id)
Table 2 HQ_ShowTypes (PKey ShowType_Id) Data(HQ_Description, end quite
some more indications)

As quite some of out companies van different systems we have a mapping of
all the local codes to the HQ codes.
This design is used in over 20 'dimensions'.

Now, the normal dimension design would be that Comp_Id, CShowType_Id if the
Primary key and that the ShowType_ID (from HQ) is used in a relation.

But in our cube we are not interested in the local id's, but the fact tables
are linking to the dimensions with these local keys.

Now there are two approaches, make dimension to the lowest level and even
hide these lowest levels in the attributes, the facts can be read easily and
easily linked. But we create quite some 'empty' cells in the cube and the
processing time of the cube might be longer.

The second approach is to eliminate the complex level in the dimension but
when preparing the fact table there, in the query do already the joins to
the hq-levels and just present the hq-levels to link.


Off course, the second one hears to be the better/faster solution, but
finally we would run this cube in a mixed MOLAP/ROLAP partition scheme ...

When in ROLAP, what would be the better solution ???


Imputs are very welcome.


(Sorry if this is a double post) but i got an error that the message was
removed from the server ....
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Top