Share via


Processing Physical Cube Partitions

Next, the partitions defined in CubePartitionInfo for each cube listed in the CubeProcInfo table are processed. The CubeProcInfo table is queried for a list of distinct ProcessingPass values. For every ProcessingPass returned, the stored procedure csdw_spGetCubeProcessingPass is called. This stored procedure returns a list of partitions associated with cubes in this pass that have the ProcessingEnabled flag set to 1.

If the Optimization value in the CubeProcInfo table associated with that partition is not NULL, then the DTS task designs and applies aggregations for the cube that target the percent optimization specified. For example, if Optimization is set to 15, then aggregations are designed with the goal of a 15% performance gain. All partitions are then processed by using the DSO processing option processFull. This rebuilds the schema and refreshes cube data.

Partitions for cubes listed in the CubeProcInfo table with the IncEnabled flag set are processed as follows:

  • If IncEnabled is set to 1, the cube is processed incrementally.

    This means if the Report preparation task is running in incremental mode, the cube is processed incrementally. Incremental processing causes a temporary partition to be created and populated with only those rows loaded into the Data Warehouse since the last time the Report preparation task ran. This temporary partition is then merged with the main cube.

  • If IncEnabled is set to 0, data for the cube's partitions is always fully refreshed when the Report preparation task is run.

In Refresh Data or Incremental mode, external processes, including the Business Desk Reports module, should be able to query a cube while it is being processed. External processes cannot query the cube during a Full process until the cube is processed. This is because the full-dimension rebuild invalidates the existing cube.

Processing Keys

The ProcessingKey is used by incremental cube processing to determine if new data has been added to a fact table but not yet loaded into a cube. New rows inserted into the ProcessingKeySource and cube fact table must have ProcessingKey values greater than existing ProcessingKey values. The ProcessingKeySource can be the fact table or another table that contains a superset of key values contained in the fact table.

See Also

ETL Process for the Report Preparation DTS Task

Copyright © 2005 Microsoft Corporation.
All rights reserved.