Changing Existing Partitions

Auto-partitioning is accomplished in two phases. In the first, the Report preparation task runs the csdw_spDefineOLAPPartitions stored procedure to create new partition definitions within the CubePartitionInfo table. In the second, the Report preparation task enumerates the defined partitions by using the csdw_spEnumOLAPPartitions stored procedure and performs the necessary actions to cause the OLAP database to conform to this partitioning schema.

The simplest method for changing the partitioning schema for a cube is to delete the entries for that cube from the CubePartitionInfo table and let the Report preparation task redefine them the next time it is run. New partition definitions will be created spanning all PartitioningKey values in the PartitioningKeySource table with each partition spanning a range of PartitionSize values.

It is also possible to manually define partitions by inserting rows directly into the CubePartitionInfo table. Note that the MinVal and MaxVal must match the data type of the partitioning key and that LastKeyProcessed must match the data type of the processing key.

There are a variety of reasons for repartitioning. One reason for repartitioning is if you delete a large amount of data from the Data Warehouse, causing some of the partitions to be empty.

Note that if you make changes to a partition such that data should be reloaded for the partition, you should either run the Report preparation DTS task in Full or Refresh modes, or change the value in CubeProcInfo.LastKeyProcessed for the partition to NULL.

Copyright © 2005 Microsoft Corporation.
All rights reserved.