cancel
Showing results for 
Search instead for 
Did you mean: 

InfoSources, transformations and performance

Former Member
0 Kudos

Hi

The scenario is: Load from DSO to Cube. The Cube has KFs that are not in the DSO and we want to have all logic in the start and/or end routine.

A solution to the problem could be to add an InfoSource in between the DSO and the Cube the InfoSource should have all fields from both DSO and Cube. The down side to the solution is that there will be an extra transformation (even though its 1:1) between the DSO and the InfoSource. Is there any one that have measures on what kind of performance impact a design like this would have?

With Kind Regards

Kristian App

Accepted Solutions (0)

Answers (3)

Answers (3)

Former Member
0 Kudos

Thank you all, I did not get much wiser, but got confirmed that the infosource method is quite valid.

Thanks

Kristian

Former Member
0 Kudos

In my case I have some models where I had the same issue.

For example I load a DSO with 5 Key Fields from a a datasource only providing 2. In that case I placed an Infosource and I created a start routine in the transformation between infosource and target DSO. This routine is distributing the results in the DSO.

It works really ok and in my case the performence is not affected. Everything depends on the routine eficiency.

mansi_dandavate
Active Contributor
0 Kudos

I guess end routine will suffice your requirement..

Why do you want to add an InfoSource in between?

Former Member
0 Kudos

Sorry but not fully OK. The end routine will only be OK if the target cube holds all the info objects that I need in my transformation. And in my case it don't. Also I would like to stay away from expert routines as they "remove" the overview.

-Kristian

mansi_dandavate
Active Contributor
0 Kudos

Ok. So I think you want to write the logic for some fields which are there in the cube based on the fields in the DSO..

Then in that case you can use Start routines along with Individual routines.

populate a global internal table in the start routine and read it in the local routine.

In 3.X there was another method to do this.

It was to add the fields in the 8<> infosource which is generated by the ODS.

But since you are using transformations you can create a infosource and go ahead with it..

There wont be any performance issues as far as you take care of it in the start routines..

Former Member
0 Kudos

Yes, I have considered the internal table solution as well. The down side here is overhead in coding and we will have code for the individual Info Objects. So what I'm after is more like the enhancement of the structure.

-Kristian

former_member205352
Active Contributor
0 Kudos

>

> The end routine will only be OK if the target cube holds all the info objects that I need in my >transformation. And in my case it don't.

> -Kristian

Can you elaborate this ?

End routine will have fields from source not target.

Edited by: Praveen G on Jun 4, 2009 10:08 AM

mansi_dandavate
Active Contributor
0 Kudos

Hi,

As I said previously, it doesnt matter whether you use the InfoSource option or the start routine option, the performance will depend on the code you write...

Let me know if you need any help in this. I know how to improve performance of a code to a good extent.