Cognos Dynamic Cube

[Cognos8]

Hi ,

We dont have separate for Dynamic cube. I am doing POC on this. I will be posting my questions with same subject and +number. If you have any idea on this question please comment. I hope that will be helpful for others as well.

1. I have created maximum of 50 dimension on single cube. to know the support number, what is the maximum dimension can be created in single dimension??? I want to compare this with transformer and Framework manager as well. let me know the maximum for this as well.
2. On single cube , one fact table is allowed to create measure. How about creting view which has multi fact table and joining it ??? how internaly it works???
3. IBM has told star and snow flake schema are best practice, then why it has given implementation tab to create join ??? just to support in worst case??
4. What is the difference between expression and specification tab ?? - as far as i know specification contains metadata ( MDX ) info. But i am not sure why it is editable.

[bus_pass_man]

1. 

Try stuff like this.   

http://www.ibm.com/developerworks/library/ba-pp-infrastructure-cognos_specific-page659/

In 10.2.2 there's a hardware sizing utility in cube designer as well.

It's not just the number of dimensions -- its the number of members too.

2.
Well, it would be just like you are making a query to the view.

You shouldn't try to fake out the query engine.  You need to take into account things like fact grains, non-conformed dimensions etc. and futzing about where you don't know what you're doing will just cause grief,  hopefully not for innocent parties.

If you have multiple fact tables, you should use virtual cubes to relate the dimensions.

In FM the dmr approach is to allow you to put a bunch of measure dimensions and their related dimensions into a package.  The scope relationships determine which dimensions will have scope to any particular fact table (via its measure dimension ) and at what level.  You can put stuff from multiple fact tables into one measure dimension in FM.  The relationships are defined in the query layer and you presumably have set the scope relationships for each measure for each level in each dimension correctly. 

If you set the scope of a dimension below the fact grain you can get values arbitrarily projected to members of the level or worse, double counting.

The approach in dynamic cubes is different.  There's a section in the modeling chapter of the redbook which deals with multifact situations.  The sample model has an illustration (not necessarily the best ) of such a situation.

3.
You might have a situation in which your data base does not have pk/fk set up.  You might need to use a view.  You'd need to have some way of joining the entities.

4.

If you look at something in the specification tab you will be things identified in this syntax <refobj> {blah}</refobj>.  The same thing in the expression tab is actually resolved to the object name that you see in the ui.   The objective is to allow you to edit complex expressions in another editor and paste them into the specification tab. 

Why just implementing a better editor was ruled out .....

The FM one was crap for long expressions too.