Measures and Multidimensional Analysis - Microsoft SQL Server 2008 Analysis Services - page 77

This can affect the performance of the system, as compared to other aggregation types
such as COUNT . In a measure group, you can have only one measure with the aggregation
function DISTINCT COUNT . If you have a measure with a DISTINCT COUNT aggregation in a
measure group, we recommend that you avoid having a measure with any other aggrega-
tion function in the same measure group.

Measure Groups

Measure groups —in other words, fact data—define the fact space of the data in the cube. In
many respects, they define the behavior of the physical data model. Facts lie on the
border between the physical and conceptual models and define the following:

. What data will be loaded into the system

. How the data will be loaded

. How the data is bound to the conceptual model of the multidimensional cube

NOTE
Don't worry if you're wondering when we decided that measure groups are the same as
facts. We introduced you to the term measure group first, but it is essentially the same
as a fact. Different uses, however, require one term or another. DDL, for example, uses
measure group .

Measures of the same granularity are united in one measure group (fact). That granularity
defines the dimensionality of the fact space and the position of the fact in the attribute
tree of each dimension.

Granularity is a characteristic of a measure group that defines its size, complexity, and
binding to the cube. In our sample cube Warehouse and Sales , we have two facts: One
contains data about sales, the other one about the warehouses. Each of these facts has a
set of dimensions, which define how the measures of a fact are bound to the cube. The
Warehouse measure group has Product , Time , Store , and Warehouse dimensions. The Sales
measure group has Product , Time , Store , Promotion , and Customer dimensions, as shown
in Figure 7.3. Each measure group can aggregate data starting from a different level. For
example, you can choose to keep daily sales data, but weekly or monthly warehouse infor-
mation. We go into greater detail about granularity in a later section, “Measure Group
Dimensions.” Having different measure groups within the same cube allows you to
analyze data related to each measure group independently and data shared between the
measure groups together in the same query. For example, you can request information
about units of a product sold in a store and units ordered from warehouse for this store.