Dimension designment

[Former Member]

Hi expert,

Is there any diference between the following cases:

1.Dimension 1

Char 1

Char 2

2. Dimension 1

char 1

Dimension 2

Char 2

for case 2,I dosen't use line item.

I think this designment will affect the perfomance,could anyone detail it,thank you in adcance,

[Former Member · Accepted solution]

Hi,

ex:

Char1 has 200 unique values.

Char2 has 300 unique values.

and if both are relevent, (for ex: Flight and Halting cities ) , then it is leading to no of records in dimnesion is 1000 (in case of 5 halts for each flight).

and if both are irrelevent, (for ex: Customer and product ) , then it can lead to no of records in dimnesion is 200*300 ( i.e maximum possiblities).

In first case we can go with one dimension with 2 characterstics.

And in second case, we better go with 2 dimnesions. one for customer and another for Material.

And of cource, some times it depends on the no of charaterstics that we take into the report.

Suppose, if we are taking only one characterstic ( either Flight or Halting cities) into report. then it is better to take 2 dimensions even in case1.Because fetching will be fast.

Regards,

Anil Kumar Sharma .P

[Former Member]

Hi,

The desigment will improve performance here.

We can create as many as 13 dimensions and it is recommended that to create as many dimensions as we can.

So Case 2 will improve your performance if data size is from medium to large.

For a scenario wih small data set, bot the cases will perform equally well but for larger data sets Case 2 will be better.

Hope it helps.

Regards,

Hardik

[Former Member]

Hi,

SAP recommends having as many dimensions as possible within the limit of 13 that we have. Creating a dimension with 10 chars gives less performance than creating 10 dimensions for the chars.

We model dimensions based on the entity relationship. Strong entities should not be put in one dimension...like customer and material which have a n:m relationship as these generate more entries in the dim tables.

In your first scenario, if you have 2 values for char1 and 4 values for char2, your dimension table could have 8 entries with the combinations.

For the second scenario, you shall have 2 entries in Dim1 and 4 entries in Dim2. This way the dim tables will be small.

I guess there would not be any difference between scenario 1 and 2 in case of chars which have 1:n relationship....like material and material type.

Edited by: Murali Krishna K on Mar 3, 2011 5:33 PM

[FCI]

To make it (very) simple, performances are directly linked to the size of your dimension tables.

If the size of your dimensions decreases between your scenario 1 and 2 then the scenario 2 should be more performant.

Regards,

Fred