BPC NW - RUNLOGIC_PH Run Time

[Former Member]

Hello All,

One of my BPC logic script is taking more than 6 hrs in prod & 30 mins in QA. To take advantage of parallel processing technique , I imported transport for RUNLOGIC_PH & created new logic script for RUNLOGIC_PH BAdI in which i called my long running script.

Changing Parameter is set only TIME dimension & DIMENSION parameter is scoped with three dimensions ENTITY , TIME & CATEGORY which are passed from DM package.

Number of parallel process are set to 4 both at Model & Environment level.

As soon as i run my DM package for one entity , one category & twelve months , i see one batch job & four dialog process running as expected !!!

But overall run time of RUNLOGIC_PH script is more than 35 mins which more than my long running script with parallel processing.

Anyone knows where parallel processing is going for toss? Why RUNLOGIC_PH with 4 processes is taking more time than single thread script?

Let me know if you need additional information.

With Regards

Abhi

[Former Member]

Hi Abhi,

Can you please attach log of that run? It should show when each process started and when it ended.

Regards,

Gersh

[Former Member]

Hello Gresh,

Log files are more than 1MB , system is not allowing to attach. Is there any other way to share with you?

[former_member186338]

First - show scripts!

Then in the logs you can remove data records.

[Former Member]

Can you zip them?

[Former Member]

please find attached files.please download and change extension to .zip. I was not able to upload zip file as sdn was throwing me error.

With Regards

Abhi

Logs_For_RUNLOGIC_PH.txt.zip

[Former Member]

I downloaded the file as Logs_For_RUNLOGIC_PH.zip then unzipped it. It shows garbage.

Was I supposed to do something different with it?

It's 585 KB packed and same size unpacked.

[former_member186338]

I was able to read the document (it's not the garbage, it's another zip file )

The script is very complex and it's hard to understand is it possible to run it in parallel without locks...

[Former Member]

Yes, it's a zip in a zip. Not sure why this was required.

Did you guess which of the attachments is the DM Package log?

[Former Member]

On page 4 . There is log file of script executed on conventional way .On page 7 , there is zip file embedded with name RUNLOGIC_PH . This file nine log file which were generated we RUNLOGIC_PH script was executed ( one for each dialog process).

With Regards

Abhi

[Former Member]

Yes scripts are indeed complex with multiple BAdI calls. Plus i am not writer of those scripts . I am just try to fine tune it & first thing that came to my mind is to try RUNLOGIC_PH to split processes in parallel.As worst case , we are running original script for each period in parallel. But issue is each DM process consumes , one batch job in NW . Users tend to run 12 jobs at time which consumes significant amount of resources(limited batch jobs) which causes other BPC jobs to go on released state.

[former_member186338]

It's not possible to help you without full understanding of the script you are trying to run in parallel! RUNLOGIC_PH is not a magic tool that will improve performance of any script

[Former Member]

Hi Abhi,

It looks like you ran different scripts in RUNLOGIC_PH and as a stand alone script.

First thing I did is looked at RUNLOGIC_PH log. It shows that periods 1 - 4 ran around 1.5 min each. Starting from period 5 each period took around 18.5 min. I don't now if it's because you have much less data in those 4 periods or because some other job started running around 13:51 and slowed this one down.

Second, I can see that all periods 1 - 4 started at exactly same time. This means that you have DEBUG = OFF in your call. Therefore those 4 logs override each other and are not reliable. This is the reason you have 9 logs and not 12. I'd recommend changing it to DEBUG =ON for the time you tune the script. You'll loose only 12 seconds:).

Third, I compared log of stand alone run with your last log RUNLOGIC_PH(9) because first logs are unreliable for the reason stated above. At the beginning everything goes as planned: RUNLOGIC retrieves 8 times less records and it takes less time. No records written back. But next, both retrieve same number of records as before and it takes same time for stand alone process and 7.5 times longer for RUNLOGIC. I'd assume that some other job started running at that time or your server was at capacity. RUNLOGIC also writes 8 times less records in twice more time.

Now BADI ZCALC_LCTOIC is called and it takes stand alone run 1.5 times longer that a single period, but it doesn't write any data back while RUNLOGIC writes 20K records back. That's where you lose 5 min.

Next RUNLOGIC call ZCALCBADI; In stand alone I don't see such call. So, there is no point analyzing it further.

Can you please find out why you have those discrepancies? Then if RUNLOGIC still doesn't run faster please send just log of DM Package, 12 logs for each period and stand alone log.

Regards,

Gersh

[former_member186338]

Please show your scripts (both)!