MSAS - Merging Partitions

Cubes with multiple partitions can be extremely confusing. Users may like to organize the cubes into a cube with fewer partitions by merging some of the partitions together. The preconditions to the merger of cubes is that the Storage and aggregation design should be the same in the partitions being merged.

To merge partitions with different storage designs, the partitions being merged must be edited to have the same storage design. The cubes must then be reprocessed before the merger is attempted. We will merge the Sales 97 partition back into the sales. Right click the Sales 97 partition and click Merge.

The screen prompting the user to select the partition for merging appears as under. Since the cube only has two partitions the other partition is displayed in the Partition list box. Select the partition and click Merge.  

The partition merge begins and the process dialog box is displayed.

Note that if the Sales partition had used a slice to limit the rows in the fact table whereas the Sales 97 partition used a filter to limit the rows of the fact table, the merger would result in the loss of the slice and the retention of the filter. After the merger the values of the cube will match the values in the fact table. On processing the cube the values of the filter will be excluded. The partition should be edited to remove the slice definition before merger, if the user wants to retain all the values.

Right click the partition Sales and click Process. Click the full process option and click Ok. Close the process log window, and browse Sales Partitioned cube data.

Now Right Click sales partition and navigate to the finish screen of the Wizard and click Advanced. Clear the contents of the filter statement box and Click design aggregations Later Option(to create a cube with no aggregations), select Process The Partition When Finished Check box and click Finish.

Click Yes when informed about counting fact table rows and close the process log window. Then browse the cube.

Note that when partitions are defined users can specify a separate data source for each partition or the user can specify a different data slice for each partition or a different data filter for each partition. A combination is also possible.

While merging partitions that have different data sources the data source for the target partition is retained and that of the merged partition is discarded. Analysis services provides no means of combining the two fact tables. The combination must be done outside Analysis services.

When two partitions using filters are merged, Analysis combines the filters using the OR expression. When two partitions using slices are merged, Analysis services, behaves peculiarly if the two slices come from two different data sources—both slices are discarded. If the slices come from the same dimension the slice of the resulting cube becomes the lowest common parent.

Therefore, before merging partitions it is important to check the data sources, data slices and data filter specifications and also that of the resulting partition.

In this lesson we have learnt all about partitions, how to work with partition wizard to create partitions, use advanced settings to create filters and how to merge partitions. To learn more about ”Implementing Calculations using MDX” navigate to the next lesson of this series.