How to improve POWL performance?

[Former Member]

I’m trying to improve performance for a Power List. Having made ​​all the necessary code improvements done, performance is still bad. I know getting a Maximum No. of Hits would help a lot but it’s not an option for the business. They wanted at least 2000 records to search.

Can anybody explain me how POWL uses data storage? Anything related to cache and PXA would be great.

Thanks!

Note: All my references are in How to develop POWL document.

[Former Member]

I'm not sure I understand. Where does the poor performance occur? In retrieving the data? Then it must be with the feeder class implementation which I assume is custom? I can display 30000 materials in POWL from MARA in less then 3 seconds.

[Former Member]

The low performance occurs mainly when retrieve data with a service manager:

lo_tor_srvmgr->query(

EXPORTING
         iv_query_key            = root_elements
         iv_fill_data                 = abap_true
         it_selection_parameters = lt_sel_param
       IMPORTING
         et_key                  = lt_key
         et_data                 = lt_fo_data.

).

I checked and we're getting about 50,000 registers. Just for this query is taking 2 minutes.

The rest of method get_objects takes a lot more.

[Former Member]

Then it is a matter of improving the performance of the data retrieval and this question belongs to the application specific space. Search SAP notes for the feeder class name and referenced ABAP classes, methods and function modules within. There is some overhead in the Web Dynpro ABAP implementation of POWL as well as the caching of queries but I don't think they are relevant here.

[Former Member]

Ok, I'll check that first.

But can you tell me more about this?

"There is some overhead in the Web Dynpro ABAP implementation of POWL as well as the caching of queries but I don't think they are relevant here."

[former_member184455]

Hello Mariana,

Can you provide a SQL-Trace (transaction ST05) of the query execution?

Is it a custom implementation of the feeder class?

What selection criteria do you apply?

If the query has no reasonable restriction on some subset (for example on 'new' documents) it may happen that despite a given maximum number of displayed documents, the whole dataset has to be analyzed. This can happen, if the query uses a SELECT DISTINCT or SELECT ... ORDER BY statement. 

Best Regards, Randolf