MSAS - Processing Dimensions and Cubes Page - 3

Processing Cubes

While processing a cube the dimension tables are read to populate the levels with members from the actual data from the tables. The fact table is read and specified aggregations are calculated. The results are stored in the cube and the cube is ready to be queried.

Three kinds of options are available for processing cubes.

Full Process: is the processing option used to perform a complete load of the cube. All dimension and fact table data is read and all specified aggregations are calculated. The cube must be processed with the Full Process option when its structure is new or when the cube, its dimensions, or its measures have undergone structural changes. In addition, virtual and linked cubes also require complete processing after they are built, or there is a change in their structure, or a change in one of their shared dimensions. To process a virtual cube or a linked cube, use the Process dialog box.

Cubes with changing dimensions are exceptions to the rule. Where only changes have been made to the structure of changing dimensions of a cube the cube need not be processed with the Full Process option. However, processing with the other options may be required.

Processing a cube with the Full Process option can take a substantial amount of time if there is a large fact table and there are many dimensions with many levels and many items in each level..

If there are changes in the data warehouse schema that affect the structure of cubes, the structure of those cubes will have to be changed and then processed with the Full Process option. If there are changes in or additions to data in the data warehouse, completely process cubes will not be required. Such changes can be incorporated into existing cubes using the Incremental update or Refresh data processing options, depending on how the data changed.

The Full Process option can be used while users continue to query a previously processed cube; however, after processing has completed, users need to disconnect and reconnect to reestablish access to the cube.

Please note that if a shared dimension's structure is updated and saved but not processed, it will be processed automatically when any cube that incorporates the dimension is processed using the Full Process option. Cubes that use this dimension will not be able to access the cube during this period.

Incremental update is used when new data is to be added to a cube, but existing data has not changed and the cube structure remains the same. The Incremental update option adds new data and updates aggregations.

An incremental update does not impact the existing data that has already been processed. It usually requires significantly less time than processing with the Full Process option. An incremental update can be performed while users continue to query the cube; after the update is complete, users have access to the additional data without having to disconnect and reconnect.

Incremental Updates and Partitions
When partitions are created and managed in multiple partition cubes, the user must take special precautions to ensure accurate cube data. Although these precautions do not usually apply to single-partition cubes, they do apply when they are incrementally updated.

On incrementally updating a cube, a new partition is created and merged with the existing partitions. The Incremental Update Wizard requires the specification of the data source and fact table of the temporary partition. It also requires the specification of a filter to limit the contents of the temporary partition. If the cube contains multiple partitions, specification of the partition into which the temporary partition is merged has to be indicated. If the cube contains only one partition, the temporary partition is merged into that partition.

To ensure accurate cube data, before performing an incremental update on any cube the user must understand the special precautions related to data integrity that apply to multiple-partition cubes. We will be discussing this in greater detail in the lesson on “Managing Partitions”

Refresh data option causes a cube's data to be cleared and reloaded and its aggregations recalculated. This option is appropriate when the underlying data in the data warehouse has changed but the cube's structure remains the same.

The Refresh data option can be performed while users continue to query the cube; after the refresh has completed, users have access to the updated data without having to disconnect and reconnect.

Additionally users can make use of a fourth option of “Incrementally update the Dimensions of a cube” in conjunction with any of the options listed above. This option incrementally updates the cubes dimensions as part of cube processing.