APD - Query as Datasource

[Former Member]

Hello,

We have an APD, that uses Query as a Data source. The query is built on Inventory Inforpovider and contains the Non-cumulative KYFs.

We have used partitioning due to large result set returned by the query.

This APD seems to be executing for very long time (over 15 hours)- we haven't seen a successful end to this run.

However, when tried to execute to one or two plants, the execution time is quite less(less than 10 mins).

Any suggestions to reduce the runtimes. we are currently on SAP 7.4 BW on HANA.

[Former Member · Accepted solution]

Hello,

Thanks for the replies. However, none actually helped and had to take help from SAP.

SAP pointed it to the UOM conversion used in BEx. and advised some Formula & Exception aggregation related notes.

We got the conversion removed from BEx and that did the magic.

[ccc_ccc]

Hi Sunny,

Could you please uncheck APD->Goto->Performance Settings-->Uncheck (by default it will be checked).

Try this

Thank you,

Nanda

[Former Member]

Hi,

have the cube been migrated to inmemory cube ? is the cube regularly compressed ? are you using partitioning by time? for non cumulative cubes you sould not use partitioning by time use semantic partitioning instead.

for more details take a look at

hope it helps

[Former Member]

Hi Anass,

The cube is an In-memory cube. The cube is not compressed - since we are in QA and have started with our initial loads.

By Partitioning I meant, using extended settings in APD.

We had referred the document. Thanks!

Cubes have been loaded fine. our issue lies with the APD runtime. Its taking more than 24 hours.to complete for a single month.

Our further analysis referred to the process getting stuck at CL_RRMS_APPL_LOG==============CP in SM50. we could see all the records loaded in Direct-Update DSO in less than an hour but the process gets stuck at this process and takes more than 24 hours to complete.

[Former Member]

The data extraction being performed here is not directly hitting the Cube DB. The query layer takes a long time to process(seek;aggregate and format).

I would suggest an alternative.

Move from CUBE to DSO(WO) Let the  aggregations and computations performed using HANA in the transformations.The DTP can be used to narrow and filter the data volume.

I am not sure what is you target , you can automate this with the PChain and APD and improve the overall performance.

Gopi