6 Replies Latest reply: Jan 8, 2015 6:00 AM by Tresesco B RSS

    Memory Friendly aggregation

    Staffan Johansson



      I have a question not a problem, I use data from a datawarehouse. In some fact tables we also use "dimensiondata" as fields.

      For example, a have 1 million rows in my fact and on every row there is a errorcode. I now want to use errorcode as my dimension in my pivot and calculate a measure from my fact table. No the question, is it better to keep the errorcode as an own dimension in the model or does this matter when I calculate this in my object (for ex pivot table)?

        • Re: Memory Friendly aggregation
          Staffan Johansson

          Just to add some information, lets say that I only have 5 different errorcode and want to sum on these 5 codes.

          If I have the dimension from fact I will have 1 Million rows as my dimension, if I instead rebuild so I have errorcode as a dimensiontable in my model I have 5 rows as my dimension.

            • Re: Memory Friendly aggregation
              Tresesco B

              Not sure if understood right. Is it something like (5 rows X 1 M cols) Vs (1 M rows X 5 cols)? If so, it would be better to convert the data in the script using probably Crosstable statement. That would not only improve the front-end performance; would simplify the expressions to get your desired result as well.

                   If you do the transformation at the back-end, it would increase the reload time which is one time job before it gets to be refresed, while for every selection in the front-end memory performance indicator would get a hit. Hence, it would probably be better to take transformation to the script.