cancel
Showing results for 
Search instead for 
Did you mean: 

Snap Shot Extractor and Cubes

Former Member
0 Kudos

Hi BW Guru's,

I have to ask your guidance on a process I have set up in BW...

My ABAP team have developed a set of 'snap shotting' extractors which extract data from R/3 on a weekly or monthly basis- this depends on what I select to do. I merely put the week ending or month ending date on the selection criteria of the info package (Depending on if I am doing a month end upload or a weekly upload) and the uploaded data is up until that date.

The reason we have done this is because we do not want to lose any R/3 history in BW every time a BW upload is carried out. We want to be able to look at data such as inventory data and see how it moves in and around different movement types over weeks and months.

I have two key fields: Fiscal Year/Period for the month and Period to analyze- week. Now I select one to kick off the upload as I stated. When the data is selected to a monthly cut off date (i.e. 06.2006) all data to this point come in with the monthly cut of date populated in the table of date. When I select by period to analyze- week this date is populated and the Fiscal Year/Period is now blank.

I am trying to capture this data on a weekly and monthly basis so I can piece the uploads together to monitor historical data with present data. Now my question is do I hold this data in one cube or do I move it from here into an export datasource fed cube. Maybe one cube can hold the monthly data and the other the weekly data. All the data comes into the PSA and is recorded depending on whether it is Monthly or weekly uploads. This way if any errors in my upload order etc are made I can blow cube data and upload from PSA.

Can anyone tell me if this methodology makes sense and what the final steps of holding this extracted data in weekly and monthly snap shots would be?

Many Thanks BW Gurus!

Depesh

Accepted Solutions (0)

Answers (1)

Answers (1)

Former Member
0 Kudos

The requirements make sense and if you cannot use non-cumulative key figures to do the snapshots your way is perfectly fine. Personally I would use two cubes especially if a) the data volume is very high or b) you have different logic depending on the time char. If you split the data into two cubes you don't need special aggregates for the different time reportings, etc. You need to duplicate the queries anyway or you create a multiprovider joining both cubes.

Best regards

Dirk

Former Member
0 Kudos

Hi Dirk,

So I will upload Weekly and Monthly data into the first cube. The dates will be populated depending on whether I have requested a weekly input date (024.2006) or a monthly (06.2006 for june). Will there not be duplicate records though or will the dates which act as a key for the result row account for this and bring up data in the query depending on which date I use in the query?

Many Thanks

Depesh