Installing SSAS and Preparing for Cube Development - Getting Started with SQL Server 2012 Cube Development - page 21

more aggregations to the cube both increases processing time as well as how much
storage is used by the cube, so the developer needs to be careful when defining the
aggregations, and define it on the levels that the users query the most.
In many cases, the in-memory model can be much quicker than the multidimensional
model. This is especially true when users often query a multidimensional model on a
level that has no aggregates. In other cases, the multidimensional cube is preferred
for performance reasons; for example, when your users only query the cubes on ag-
gregated values, orifyouhavealargeamountofdatathat willnotfit intothememory
of your server.
Custom assemblies
In the multidimensional model, you have the possibility to run custom assemblies.
Custom assemblies are a way to extend the functionality of Analysis Services. You
can write .NET programs that can perform a specialized task that will be loaded into
and executed by Analysis Services. An example of common custom assemblies are
specific security models that cannot be implemented easily with the normal security
mechanism in Analysis Services, but that can be solved using .NET code.
Custom rollups
In cubes you have the possibility of defining measures, which is a calculation on a
value. Most often they use common formulas such as sum, average, or count. In
some cases, you may want to add custom rollups that change the way a measure
is calculated on a specific level. As an example, think of a measure summarizing a
value over a time dimension; for business reasons, you may want to have a specif-
ic formula for the year level and not a summary of all the months or quarters. Cus-
tom rollups allows you to write such logic and they only exist in the multidimensional
world.
Distinct count
In multidimensional cubes, you have the possibility of creating distinct count meas-
ures. This special type of measure counts the distinct values instead of counting all
the values. In the tabular world, you have to define this using a simple DAX for-
mula instead. The formula is very simple and looks like the following: [Measure] =