COGNOiSe.com - The IBM Cognos Community

IBM Cognos 8 Platform => COGNOS 8 => Framework Manager => Topic started by: saumil287 on 21 Sep 2011 12:58:35 AM

Title: Multi Dimensional Model
Post by: saumil287 on 21 Sep 2011 12:58:35 AM
Hi all,
I want to know whether Star Schema Grouping is Mandatory in order to create a multidimensional package in cognos fm 8.4.1 for Analysis studio.

Regards
saumil
Title: Re: Multi Dimensional Model
Post by: blom0344 on 21 Sep 2011 02:13:03 AM
It is within the roam of 'best practice', but if you have cardinalities set up properly even a snowflake can yield a proper dimensional model. Star schema grouping is mainly a wizard to make the modellers life a little easier. Actually never used it and still build a fair number of dimensional models..
Title: Re: Multi Dimensional Model
Post by: saumil287 on 21 Sep 2011 02:25:02 AM
Thank You for your reply sir,
one more thing I would like to know in general .
What other compainies follows when building a model.
ie Do they build a star schema or if the Reporting requirements are met without creating a star schema model , They publishes it
I am very much confused while creating a model.
Thanks in advance
saumil

Title: Re: Multi Dimensional Model
Post by: blom0344 on 21 Sep 2011 05:28:13 AM
A star schema would be closest to the ideal. Nevertheless, there are truly relational models that will not fit into the perfect picture. If you build a dimensional model for Analysis Studio then any combination of dimension and measure should either yield correct results , or when out of scope then yield no result.  The phrase 'in general' is really the problem. There is no 1 standard solution that fits all
Title: Re: Multi Dimensional Model
Post by: saumil287 on 22 Sep 2011 05:09:50 AM
Thank yopu for your reply sir,
I want to know that how can we decide whether our model needs to be a star schema or snowflaked schema or none , if the reporting needs are met.
Title: Re: Multi Dimensional Model
Post by: blom0344 on 22 Sep 2011 01:47:22 PM
No No.  8)  You need to approach this the other way around. The closer you can create a model as (a collection of) starschema's, the better. However, this depends on lots of variables. If your source is no DWH / datamart then you may have to compromise. However, cognos framework allows you to 'bend' the underlying database quite a bit towards the 'ideal'. Unfortunately this is not something I can give direct advice about without knowing what you are up against   ;)