Multidimensional Clustering (MDC) for InfoCubes, DataStore Objects and PSA Tables (New) 

Technical Data

Function is	New
In Release	Software Component ●      Component: SAP NetWeaver ●      Release 7.0
Assignment to Application Component	BW-SYS-DB BW database platforms
Country Setting	Valid for all countries

Use

With SAP NetWeaver 7.0, multidimensional clustering is supported for InfoCubes and DataStore objects.

The performance of database operations can be increased markedly using multidimensional clustering. In multidimensional clustering, the data records of a table are sorted by one or more dimensions and saved in the database. This increases read-, write- and delete access to the table.

For fact tables of an InfoCube and its aggregates, one or more dimensions can now be selected for multidimensional clustering. Clustering can be set before you activate an InfoCube in the Extras ® Database Performance ® Clustering menu.

For the active table of a DataStore object as well, one or more InfoObjects can now be selected for multidimensional clustering. Clustering can be set before you activate a DataStore object in the Extras ® Database Performance ® Clustering menu.

PSA tables and PSA-like tables from DataStore objects are now created by default with multidimensional clustering on the REQUEST field.

Multidimensional clustering is currently only being supported by the database platform IBM DB3 Universal Database for UNIX and Windows.

Effects on Existing Data

In order to be able to use multidimensional clustering for an existing table, the table has to be converted. For fact tables from InfoCubes and for active tables from DataStore objects, the new Reclustering function is available.

For more information, see Repartitioning InfoCubes and DataStore Objects (New).

Effects on System Administration

For multidimensional clustering, a small extentsize of the tablespace is recommended. Especially for fact tables for InfoCubes and aggregates, the previous, default extentsize of the tablespace of 16 pages belonging to data class DFACT is too large. After an upgrade, it makes sense to create a new tablespace with an extentsize of two pages (minimum value) and to put InfoCubes with multidimensional clustering in the new data class. In newly installed systems, the tablespace of the data class DFACT already has an extent size of two pages.

See also:

Extensive documentation about multidimensional clustering can be found in the SAP Library under Clustering.

For more comprehensive information and examples for using multidimensional clustering in BI, see the document “SAP NetWeaver Business Intelligence – Database Administration Tasks: IBM DB2 Universal Database for UNIX and Windows”, located in the SAP Service Marketplace.

The manual on the database IBM DB2 Universal Database for UNIX and Windows “Administration Guide: Planning”) includes a technical description of multidimensional clustering.