Live cache to APO BW timing issues

[Former Member]

Hello,

I have a question regarding the time it takes to copy data from Live Cache (LC) to APO BW. Say we upload 1000 CVCs (10,000 records) from flatfile into LC directly (takes less than 2 minutes) and copy this data into APO BW.Currently there are 200,000 CVCs and it takes around 50 minutes to copy from LC to APO BW. Will there be any reduction in the time from LC to APO BW if we reduce the number of CVC s from 180000 to 50%, 75% (90,000 or 45,000).

No operation(drill down, macros) are done in APO currently. We need to only know if LC to APO BW has any impact on CVC reduction.

Really appreciate your inputs.

Thanks,

Sai.

[Former Member · Accepted solution]

Hi

CVC reduction will not help in reduction in time.

Request your basis team to put a trace on the transaction. This will give you the split for various activities like direct read,sequential read etc and you may get the bottleneck activity.

Parallel processing profile can be useful.

In case you are already using it, please check how many parallel processes are getting invoked during this activity. The greater the number the lesser the time.

Also if there are parallel jobs running in the system, it can lead to resource crunch.

Thanks

Amit

[former_member209769]

Hi Sai,

I could not understand what you mean when you say that you copy data from APO livecache to APO BW.

No of CVCs would be directly linked to the time that would be needed for any kind of data transfer. The more the data, the more time it would take and vice-versa.

regards- Pawan

[former_member187488]

Hello,

I also think number of CVCs will affect the processing time.

I suppose what Sai is doing is extracting data to BW by running infopackage ....

The processing time of this process should be relevant to the below settings or contents:

1)Number of CVCs to be processed.

2)Time series (planning period) length.

3)Whether parallel processing profile is specified in the datasource.

4)You can spcify not to extract zero records in the datasource.

5)If data is read from liveCache on aggregated level, fixed aggregates may help to improve the performance of reading from liveCache.

Maybe there's still something else

Best Regards,

Ada

[Former Member]

Thank you everyone for the responses. But the requirement here is like the following.

Every week there will multiple uploads of same flat file (different versions) into LC. Before every upload to LC the previous data will be deleted. Say i upload 10000 records into 1000 CVCs. When i upload again i clean up the t data from previous load and upload again.

But the time consuming part is LC to APO BW which is part of the process before the bex report is generated from BW (LC to APO BW and then APO BW to BW) . So what we want to avoid is to avoid this time consuming process (LC to APO BW) for every load as there is one single version which is finally approved.

This is the business process so we dont have much control or visibility on why they upload so many versions in single week.

So coming back to number of CVCs all the way i was under impression that every time LC to APO BW extract happens , i thought all the CVCs will be extracted. But only yesterday i came to know that before every upload the LC is totally cleaned up for next upload. So i guess it is not the problem of number of CVCs (Because they are constant for every upload) but the LC to extraction in to remote cubes using Bapis that is taking time. Does any one know what is the general time ( 50,000 cvc s ) to extract from LC to APO BW? In our case it is little over 40 minutes.

[Former Member]

Hi SAPSCM,

1. when you say LC to APO BW do you store data in cube and than copy to BW? If yes, you do not need to load data to APO BW. you can create data source and pull data directly to BW. This should save you some time.

2. Other thing is, if you have enough horse power and you load BQ just to report out of it, in APO than you can directly create BEX report on APO. This will eliminate need to load data into BW.

3. When you load data into APO BW how many threads run in parallel? you can also run multiple threads and improve performance.

Hope this helps.

Regards,

Harshil Desai

[Former Member]

Hello Harshil,

The 2 point is useful. I was under impression that it is not recommended to create Bex query directly on APO BW. So when u say enough horsepower how many times of current system are we talking about (1.5X,2X ...?)

Just to make sure when u say APO it is APO BW right? not LC?

So but the point here is APO to BW copy also does not take much time. Is it normal for extract from LC to APO BW to take long time?

[former_member209769]

Hi Sai,

So it's clear now to me that you are talking about time for extraction of planning area data from APO into BW.

The extraction normally does take quite some time. 40 mins is not huge in my understanding considering that you would also have some routines in BW extraction. (Time also depends on free processes available in APO and in BW for your data extraction)

If you want to reduce the time, then you could do the following:

1) Have 2 separate data source for APO planning area - one corresponding to history related data (time horizon in past), and the other relating to forecast horizon in future (time horizon in future). In the respective datasource, only select the Key Figures (KFs) relevant for that horizon, e.g. for past. select KFs like forecast, past promotions, history adjustments, etc. Similarly for future data source, select KFs like stat forecast, forecast adjustments, consensus forecast, etc.

In the extraction data source, you could specify to "ignore zero values". So, if your KF value is zero, that particular record would not get selected for extraction. This helps in reduction of data volume for extraction, and hence overall reduction in time in many cases.

2) Discuss with BW ABAP persons. There are some possibilities that they can optimize your BW routines for better performance using ABAP good practices.

Thanks - Pawan