create similar dimensions or use subsets?

[BigG]

Hi,

1st question leading me to the rest of the post- Does TM1 performance get affected by multiple hierarchies in a dimension?

Have a output cube that requires multiple dimensions (accounts), which is fine.
Have an input cube that could use a subset (data entry accounts- approx 80% of total accounts). This input cube doesn't need multiple hierarchies - only one rollup.

2nd Question is should I use the already available dimension and create subset based on attribute flag for data entry, or would it be more efficient to create a more specific dimension off the main 'account' dimension for data entry (with single hierarchy)? I would run this as a TI process after the administrator has manually flagged the data entry accounts (which they have to do anyway)

Any feedback would help. Thanks in advance

[lotsaram]

Does TM1 performance get affected by multiple hierarchies in a dimension?

Short answer is no. Multiple hierarchies don't really affect TM1 as non-leaf members are only ever calculated on demand.

should I use the already available dimension and create subset based on attribute flag for data entry, or would it be more efficient to create a more specific dimension off the main 'account' dimension for data entry (with single hierarchy)? I would run this as a TI process after the administrator has manually flagged the data entry accounts

There's no right answer to this question. My own approach would be to use a subset as why create an additional dimension if it is not really needed. Sometimes auxiliary dimensions can be quite useful, say when distinct element security settings are required. But in this case why bother when you can just create a subset.

[tomok]

The answer here really more lies in personal preference. When I need only a very small portion of a dimension for a different cube then my preference is to create a separate dimension in order to make the cube smaller and to reduce confusion amongst the user base (should they go beyond a subset by clicking on Show All). A perfect example of this is the GL chart of accounts. Certainly you can create a single Account dimension and put every account in it, or you can create separate Account dimensions for each functional cube. For example, when I create a capital expenditures cube I take the subset of accounts where capital expenditures and depreciation are recorded and create an Account_CapEx dimension. That's because those accounts are such a small section of the overall chart. In your example, your second dimension is 80% of the original dimension so I would probably not create a second dimension. Once again, it's personal preference and your appetite for maintenance (more dimensions, more maintenance).