All data are stored in logical sets of data. As a minimum one partition will be created for each single measure group in the OLAP database. Each partition may be physically stored, maintained, tuned and processed independently of other partitions.
Custom partitions, e.g. splitting of measure group partitions, may also be designed to prevent large partitions from being slow or inaccessible. For example one might create partitions of a cube for each Year of data to obtain smaller, easier maintainable and faster-querying chunks of data.
The default setting for partitions is that one partition is made for each measure group.
To define custom partitions, go to the Partitions page of the Sales cube.
On the Partitions page you can create new partitions in addition to the default partitions.
For example we might want to split the Sales Invoice Line partition into several smaller partitions, e.g. one partition for each year of data.
To do this we will first need to delete the existing default partition.
Now we click New Partition to create a Partition containing data from 2012.
We specify the Fact table for this partition.
Set a checkmark and define a query that filters the data by Posting Date.
Clicking next a couple of times takes us to the finish.
Repeating the process, we create another partition to hold the old data from before 2012.
Tip: Splitting data into year-by-year partitions may in fact reduce the time needed for the daily update and processing of the cubes. Partitions holding data older than 2 years or more are probably quite static and thus do not need to be updated and processed every single day. In this way only the partition(s) with the latest data need to be updated and processed on a daily basis.