Number Range Buffering - Daily Full load, Max Value, DTP settings, Implementation for SPO

[former_member213072]

Hello folks,

I have a few Questions on Applying Number Range Buffering to SPO based Cube:

1. Daily Full Load: Our current requirement is to daily delete and load data to cube. We have 3 sources with each having around 500K-1,000K records being loaded on daily basis. With very little Routine/Logic in the transformation, currently takes around 3-4 Hours - which is really too high(Since we already have lot of coding at Data source level and we require a DSO between datasource and Cube, which already takes some time).

Question :a. If we are performing Daily Full load, can we set the Number Range Buffering to a max value and leave it permanently?

               b. If we delete the contents of the cube and load data afresh will it better in performance when compared to delete overlapping requests after                      loading the new data?

2. Max value that can be passed in No. Of Numbers in Buffer: Though SAP suggests that a value between 100 to 500 would suffice most requirements.

Question :Since we are daily loading huge volumes can we set the value to 5000 or higher. Will this better the load performance when compared to                        providing 100-500.

3. Does reducing Package Size of the DTP improve the performance: When we have a Buffer set to 5000, if we set the package size of the DTP to 5000 will it provide performance than selecting package size as 50,000. Will there be any major difference when we reduce the DTP package size to the Buffer store value?

4. Implementation of Number range Buffering for a SPO based on Cube: Though the data would exist in the Individual partition Cubes, would just applying Number range buffering to the required dimensions of individual Cubes be sufficient or do we need to consider any other points specific to SPO.

Please let me know, if you need more information.

Regards,

Thejas K

[prakashsingh14]

HI Thejas

Setting a huge Number range buffering wont have much impact on loading performance. Are you deleting the indexes before loading data to cube. Also as you dont have complex routines in transformation, reducing data package wont be helpful. Check the number of back groung process assigned to DTP, check how many BG process are free in sm50 and assign accordingly.

Thanks

Prakash