Aggregation Design and Usage-Based Optimization - Microsoft SQL Server 2008 Analysis Services - page 417

In the earlier example, suppose that there is no relationship between the Store State
and Store City attributes. Even though you can't have multiple cities with the same
name in a state, you can have a city with same name located in several states (Portland,
for example, in Oregon and Maine, to name two). In Figure 23.7, you can see that an
aggregation that contains more than a single attribute from a dimension is valid rather
than redundant.

You can calculate how many aggregations you need to maximize query performance in a
cube that has relational reporting-style dimensions. For this calculation, an aggregation
definition is represented as a series of digits, where every attribute in every dimension is
represented by either 1 or 0 . 1 represents the presence of the attribute; 0 represents its
absence in the aggregation definition. These digits are grouped according to dimensions,
with the groups separated by commas.

Brand Month Store State Store City City Units_Ordered Units_shipped

Booker February WA Tacoma Ta c o m a 47 47

Carlson February WA Seattle Seattle 86 86

FIGURE 23.6 The Store State and Store City attributes define an aggregation that is
the same as one without the Store State attribute.

Brand Month Store State Store City City Units_Ordered Units_shipped

Booker February WA Ta c o m a Ta c o m a 47 47

Carlson February OR Portland Portland 144 144

Carlson February MN Portland Portland 23 23

FIGURE 23.7

The Store State and Store City attributes define a valid aggregation.

For example, this series of digits represents an aggregation in the Warehouse measure
group: 001000,0000,0000,0000 . The first group of digits represents attributes from the
Product dimension. The digit 1 in the third position indicates that third attribute of the
dimension, Product Category , participates in the aggregation. In the next group, there is