dataflow hbmp+ or CellBased+

Hello,

I create a dataflow to divide the planed article sales on the customers. 

First cube (ZW2_Umsatz) has the dimensions month and Produktmodell (article). Second cube (percent per customer)  has only dimension customer .

If I calculate this two cubes to a new target cube (ZW3_Umsatz) with the dimension month, Produktmodell (article) and customer it takes a long time, about 13 minutes. The log file tells me it was calculated in CellBased+.

 

Now I made a test and added in my second cube (percent per customer) an additional dimension named ordermedium. But the dimension is not in my target cube! So it has no function at all. If I now calculate with the same procedure (nothing more changed) it goes with HBMP+ and takes only 10 seconds.

 

What is the reason for this?

 

Best regards

Marco

Tagged:

Answers

  • Björn Reuber
    Björn Reuber Employee, Community Captain
    500 Likes Fourth Anniversary 100 Comments 5 Answers
    edited March 2020

    Hi Marco,

     

    from my understanding the 2nd testcase shouldn't be HBMP either, cause there are some requirements missing.

     

    Have you tried the frist DF only with High Performance Mode, or also with Try Force HBMP mode? image

    To get this option you have to tick the option for "Show layout script tab"

    image

     

    Regards

    Björn

  • Previous Member
    edited March 2020

    Hello Björn,

     

    this is the log file with my test-examples:

    image

    Red is my first testcase and blue is the second testcase.

     

    Now I tried your option. For my first testcase BOARD calculate with HBMP if I activate the option "Try Force HBMP mode" and it takes only 11 seconds.

     

    Thanks for your tip!

     

    Best regards

    Marco

  • Björn Reuber
    Björn Reuber Employee, Community Captain
    500 Likes Fourth Anniversary 100 Comments 5 Answers

    Hi Marco,

     

    keep in mind to also validate the figures, cause some calculation can't work in HPM Mode, also you can force them. So it might be a (for human) confusing result, but correct from HPM view

     

    regards

    Björn