Managing Partitions

Partitions must be created and managed correctly to avoid inconsistent or inaccurate results. This requirement applies to multiple-partition cubes. It also applies when you incrementally update any cube, including a single-partition cube, because an incremental update creates a temporary partition and merges it into an existing partition.

The integrity of a cube's data relies on the data being distributed among the partitions of the cube such that no data is duplicated among the partitions. When data is summarized from the partitions, any data elements that are present in more than one partition will be summarized as if they were different data elements. This can result in incorrect summaries and erroneous data provided to the end user. For example, if a sales transaction for Product X is duplicated in the fact tables for two partitions, summaries of Product X sales can include a double accounting of the duplicated transaction.

Partitions can be merged; you can use this feature in your overall storage and data update strategy. Partitions can be merged only if they have the same storage mode and aggregation design. To create partitions that are candidates for later merging, you can copy the aggregation design of another partition when you create partitions. You can also edit a partition after it has been created to copy the aggregation design of another partition. Merging partitions must also be performed carefully to avoid duplication of data in the resulting partition, which can cause cube data to be inaccurate.

When you are creating or merging partitions, you may need to perform manual operations on underlying dataor create appropriate filters to ensure that the partitions of the cube always contain the correct data. This topic addresses issues and precautions to be aware of when you are creating or merging partitions or performing incremental updates of cubes.

See Also

Incremental Updates and Partitions