Database Reference
In-Depth Information
Table 12-1.
The Review Aggregation Usage Options
Option
Description
Default
The attribute may qualify for aggregation. The designer chooses aggregations for you using
a Microsoft algorithm, which is based upon the type setting of attribute and dimension. The
chosen aggregation will be considered as the wizard continues.
Full
The attribute will qualify for aggregation, and all aggregations will be considered as the wizard
continues. This setting can cause excessive processing time if there are a lot of rows in the
dimension tables.
None
The attribute will not qualify for aggregation, and no aggregation will be considered as the
wizard continues.
Unrestricted
This uses a modified version of Microsoft's default algorithm, where no restrictions are put on
the aggregation designer; however, the attribute must still be evaluated to determine whether
it is a valuable aggregation candidate.
Clicking
Set All to Default
at the bottom of this window resets all of the settings for each attribute to the
default radio button. In most cases, this is an appropriate choice, because Microsoft's algorithms have proven to
be competent at choosing which aggregations are most useful.
Clicking Next brings you to the Specify Object Counts dialog window (Figure
12-21
) that contains Count and
Stop buttons. For the wizard to continue, you must let SSAS count how many rows are in each dimension table.
Clicking the Count button starts this procedure.
Usually it takes only a few minutes or less for it to count the rows, but it will take a very long time if there are
many rows in each table. As such, Microsoft has provided the Stop button in case this takes too long. But be
aware that you cannot continue through the other screens of the wizard without having created an estimated
count for each table, so stopping the count is only useful if you plan to try once again, when there is less activity
on the network. Therefore, once you click Stop, you must exit the wizard and create the aggregations at another
point in time.
Not
■
This may seem quite tragic until you realize that cube based reports work just fine without stored
aggregations. The difference is that they are just not as fast as they could be.
