7 Replies Latest reply: Jun 10, 2014 8:18 AM by TimG RSS

    Dimension to be tagged as dense?

    NJ

      Hello Essbase gurus,

      I am aware of dense and spare tagging in essbase. Dimension where probability of data is more is to be tagged as DENSE.

      My question is whether you tag a dimension as DENSE depends whether or not density of data is 100 % for that dimension.

       

      Say if I have dimension with respective % of data probability. Which dimension wud u tag as dense and which as sparse.

      A - 100%

      B - 95%

      C - 80%

        • 1. Re: Dimension to be tagged as dense?
          KKT

          Ideally Account and Period are always dense.

           

          But depending upon your business requirement and Dimesnions in your application you can change the property to either dense or sparse.

           

          Have a look at Essbase DBAG to know more.

           

          Thanks,

          ~KKT~

          • 2. Re: Dimension to be tagged as dense?
            NJ

            Thank you for the response KKK.

            I am aware of the concepts here. The question is more on generic lines. You can leave aside Accounts and period Dimension in this context.

            • 3. Re: Dimension to be tagged as dense?
              KKT

              As per your example I will Tag A,B and C as Dense Only as it will lead to data blocks and page file calculations and performance impact as well.

               

              To explain and you are aware  - All of the dimensions are taken into consideration when creating a data  value, both sparse and dense. You can't get around this and in Essbase you often see "No Product" or "No Customer" to allow the storage of a

              data point that doesn't really make sense by Product or Customer.

               

              Thanks,

              ~KKT~

              • 4. Re: Dimension to be tagged as dense?
                Phulsebosch

                Hello NJ,

                 

                the stored members of the dense dimensions are building the data block. A data block should not be too small and not be too large. Therefore, a large dimension will not likely be a candidate for becoming dense. Then I would prefer 3 small ones, even if they have a less data probability. When there are member formulas, or I can use member formulas instead of bringing them into a calc script, then this increases the possibility of this dimension becoming dense.

                 

                The given percentages of data probability are very high. I don't know how you got to these. I usually imagine a given data set with x thousand records. How many blocks would be created out of these, so the (remaining) sparse dimensions determine the data probability.

                 

                With other words, you cannot decide very well from the given examples. It requires a lot more.

                 

                I hope this helps.

                Regards,

                 

                Philip Hulsebosch

                www.trexco.nl

                • 5. Re: Dimension to be tagged as dense?
                  NJ

                  Thank u for your responses !!!.

                  Much Appreciated !!

                  • 6. Re: Dimension to be tagged as dense?
                    GlennS_3

                    As Philip says, there is too little information to make a decision. Things to consider are

                    How large the dimensions re

                    how often they change (add/delete members)

                    Are formulas used in the dimensions

                    Do you need to consider block creation issues when calculating

                    etc

                    • 7. Re: Dimension to be tagged as dense?
                      TimG

                      NJ wrote:

                       

                      Say if I have dimension with respective % of data probability. Which dimension wud u tag as dense and which as sparse.

                      A - 100%

                      B - 95%

                      C - 80%

                      Hi NJ.  Can you define "% of data probability"?

                       

                      For making dense / sparse decisions, the 'density' that matters is the ratio of members populated in the candidate dimension, for a given populated intersection of all the other dimensions.  I would be very surprised (as Philip noted) if you really have a dimension that is 100% dense by this definition - unless it's an analytic dimension with only one stored member!