cancel
Showing results for 
Search instead for 
Did you mean: 

logical grouping of dimensions

Former Member
0 Kudos
60

Hi,

What is criteria to enter dimension into the cube and what does mean by logical grouping of dimensions

Thanks in advance

krish

Accepted Solutions (1)

Accepted Solutions (1)

gunesbt
Active Contributor
0 Kudos

Hi,

While designing the dimensions, you must ask to yourself that "What will be the size of dimension table if I add this characteristic to this dimension?".The smaller ratio of size of dimension table/ fact table means better performance.For this ratio, 20% is dangerous and 10% is preffered.

If the characteristics included in one dimension have mxn relationship, then your dimension table will be too big(specially if those charactersitics are already big) and performance will be decreased.Therefore, 2 characteristics that have mxn relationship(e.g sales document and sales document item) must be in seperate dimensions i.e two characteristics that have nx1 relationship(e.g material and material group) may be put into one dimension(Not logically related(groupped) characteristics, this is not a criteria)Two characteristics that have nx1 relationship in one dimension will add only m+n rows to dimension table instead of mxn.To sum up, you must create the dimensions by considering the relationships of characteristics that will be included in the cube.I propose you to activate some SAP cubes(e.g 0FIAR_C03) from BI content to view their dimensions.You can see the logic there.

You can see the size of dimension table and fact table by executing the program SAP_INFOCUBE_DESIGNS in se38.

Useful threads:

Regards,

Güneş

Answers (2)

Answers (2)

Former Member
0 Kudos

Hi,

you'll have to calculate how many records can your dimension portentially have.

This will depend on which characteristics are packed in it.

e.g. you have COMP_CODE, SALESORG, and PLANT in one DIM.

Because one plant belongs to one SALESORG which belongs to one COMP_CODE, this DIM will have the same cardinality than the PLANT (=same number of records than plants on which facts will be posted).

Now assume PLANT and MATERIAL in the same dimension of a sales cube.

You have 100'000 materials.

You have 100 plants.

Assuming that each plant can sell any material, your dimension will have "potentially" 10 mio records....

this is how we decided the grouping of infobjects in dimensions, so that fact tables doesnt effected so much & ratio of dim table to fact table should be 10% , then only it will be good in performance......

hope this helps....

ravi_chandra3
Active Contributor
0 Kudos

HI

U can place two characteristic in a single dimension if and only if the cartisean product of the 2 characteristic results in a smaller dimension table . The best example is the MATERIAL and MATERIAL GROUP .

Let's take a simple example like Electronic goods which is a material group and the and the various electronic equipment like computers , mobile etc.. and here in this the relation between them is 1 : M . Mean 1 electronic goods may comprise M no. of materials lik computers , mobile etc. and the cartiesan product between them is M ( 1 * M ) so this will results is a smaller dimension table . So in this case we can place both the characteristic in a single dimension table.

Note: Suppose if the relation between two characteristic is M * N and the cartisan product between them results in a smaller dimension table then we can place them in a single dimension else we have to place them in a separate dimension.

Hope u got it,

Thanx & Regards,

RaviChandra