Sparse VS Dense and neglecting Period Dimension

Question

Hi All,
I am working on a classic planning application, in which i have 3 custom dimensions apart from six standard dimension. As the application grows and we start developing business rule we feel the need for optimization of our application hence along with increasing Cache Size, Block Size etc. we also make changes in JVM registry and lastly we also convert our Custom dimension to Dense from Sparse. As we are always using these dimension at either row/columns. Now i have to different questions:
1) there are three custom attributes associated with one of my Custom dimension with 150descendants, so if i change it to Dense then i loss those three custom attributes. So is there any way to convert this from sparse to dense while having the attributes with it. Note: I am using this dimension on Row and i only want to convert it to dense in order to enhance performance. The motivation behind this change is that i piratically experiment that the business rules that take more then 20mins to executes just execute in less than a minute when i convert the other 2 custom dimensions from sparse to dense.
2) I am not using Period Dimension in my application, because my need are fulfiled by year dimension only, so is there any way to delete it or neglect it while processing business rules?

Answer 1

Hello,
Your question gets into some advanced design topics fairly quickly and you may want to engage an experienced consultant to help you.
It sounds like your application is past the design stage and you are making what are design level changes which is a likely risk to your project delivery time frames.
Regarding the conversion of the dimension from sparse to dense -- there is no way to tag a dense dimension as an attribute.
Review Chapter 10 of the Database Admin Guide at http://docs.oracle.com/cd/E17236_01/epm.1112/esb_dbag/frameset.htm?dattrib.html#dattrib524566
Pay special attention to the "Understanding the Rules for Base and Attribute Dimensions and Members".
In addition to loosing the attribute dimension please be aware of the implication to block size. Switching a sparse to a dense has a multiplicative effect on your block size; you should make all upper level members dynamic calc and then the number of level 0 members time your old block size will be your new block size.
On the topic of your period dimension you may want to try renaming it and using it for one of your other dimension needs if you do not require month level data.
On the topic of calculation speed you can usually make re-write your calculations to make them dense. In your instance it may be a good idea to make your years dense to use that to facilitate creating dense calculations.
Good luck,
John A. Booth
http://www.metavero.com