High consume of CONCAT_ATTRIBUTEs into FAGLFLEXA (FI view)

[Former Member]

Hi colleagues,

I had side car scenario implemented that includes de FI accelerator on HANA. I

noticed that this scenario uses two views that are generated

automatically when the FI accelerator is activated via HDBC. The views

are: V_GLPOS_C_CT and V_GLPOS_N_CT. However I see that these views are

materialized on HANA, so there was an increased of consumption (more

than 50GB). I executed the script HANA_Tables_ColumnStore_Columns

(recommended in note 1969700) - I can attach the result if you want.

Also, I found  the note 1986747 - Internal columns in Column Store and I read

that CONCAT_ATTRIBUTES was not necessary since SP07 (we are on

SP08). So I would like to know if it is possible to drop those

"concat_attributes" and create a virtual view without impacting the FI

side car.

Let me know if you have any recommendation for this case. This could

release us almost 50GB of memory/disk usage.

Thanks in advance.

Regards

Amanda

[lbreddemann]

Hey Amanda,

whether or not the concat attributes can be dropped depends on what they are used for.

Attaching the output from the script is a good idea.

- Lars

[Former Member]

Hi Lars,

Thank you! I just attached the file.

My question is whether the concat_attribute is still required for FI accelerator scenarion, or if I could drop it after the HANA SP08.

Thank so much.

Amanda

[lbreddemann]

Hi Amanda,

with SPS 9 the only type of concat attribute you can change are the ones used for information models and multi-column joins.

As of SPS 9 you could convert the concat attributes used for unique and primary key to INVERTED HASH to save storage space.

However, for both options, you will have to try out the effect of the change to your query runtime performance.

There is no general recommendation to drop the concat attributes at all. So, if you do drop them you need to perform some testing for performance and function.

Especially for concat attributes that have been created by the activation of information model views, you will have to re-activate the corresponding views.

So: no easy YES or NO answer to this one. It's more complicated than that and cannot be handled automatically.

Looking at the list you attached, I would recommend to upgrade to SPS 9 and check if the performance with INVERTED HASH unique/primary keys would be good enough, since the largest concat actually is "just" a unique index.

- Lars

[Former Member]

Great, Lars! It's very clear now. I will upgrade to SP09 and test it.

Thank you so much.

Regards,

Amanda