Point 1:
Don't use too deep levels, it's ok to have a lot of members in one level. But it is not wise if we have deep nested level, and each level doesn't have too much members. This is very important principle when we design a cube.
The outline secuqence: time, account, the other dense dimensions, the sparse dimension that has fewest members, other sparse dimensions, the attribute dimensions.
Point 2:
Calculation performance may be affected if a database outline has multiple flat dimensions. A flat dimension has very few parents, and each parent has many thousands of children; in other words, flat dimensions have many members and few levels. You can improve performance for outlines with multiple flat dimensions by adding intermediate levels to the database outline.
The above 2 points are from different source, they looks some different and even in the opporsite. my understanding is: we should have fewer levels anyway, but the huge amount of member should happen in the parent level. That means we have thousands of parents, but each parent has few members.
