|Oracle9i OLAP User's Guide
Release 2 (220.127.116.11)
Part Number A95295-02
Whether you choose to use grouping set or concatenated rollup for your fact materialized views will depend on the complexity of the data in your star schema and on the nature of your OLAP metadata.
For more information on the metadata requirements, see "Materialized Views and OLAP Metadata".
Unless you have a very simple data model with only single-hierarchy dimensions, grouping set MVs are generally more efficient and provide greater flexibility than concatenated rollup MVs.
MVs generated with grouping sets provide better runtime query performance for schemas that have dimensions with multiple hierarchies. MVs generated with concatenated rollup are more efficient for schemas that have only single-hierarchy dimensions.
If you have single-hierarchy dimensions, concatenated rollup MVs will take less time to build than grouping set MVs. If you have multiple-hierarchy dimensions, grouping set MVs generally will take less time to build.
If you want to store partially aggregated data in your materialized views, the grouping set form provides more flexibility than the concatenated rollup form. Grouping set form supports asymmetric partial materialization. Concatenated rollup form supports only symmetric partial materialization.
With grouping set form, you could store month level summaries for specific level combinations only. For example, you could summarize month data for a certain type of product within a given geographical region, without regard for the other dimension levels associated with the data. You would do this by specifying individual level combinations before generating the script for creating the MV.
With concatenated rollup form, you could store month level summaries only, but they would be aggregated over all of the dimension hierarchies associated with the cube. You could choose to limit the MV to month data by editing the script for creating the MV.
Although a grouping set style MV may be very large, it requires significantly less tablespace than concatenated rollup style MVs. The multiple concatenated rollup style MVs for a cube store redundant data, since each hierarchy combination is stored in a separate MV. A grouping set style MV for a cube contains all hierarchy combinations within the single MV.
With concatenated rollup form, all the dimension key columns are populated, and data may only be accessed when its full lineage is specified. With true grouping set form, dimension key columns may contain null values, and data may be accessed simply by specifying one or more levels.
In the current release, all MVs, whether generated with concatenated rollup or with grouping sets, are full lineage preserving.