Diagnosis
You want to select InfoCube dimension for multidimensional clustering.

System Response
Multidimensional Clustering (MDC) organizes a table physically intoblocks. Each block contains only data records with the same values inthe MDC dimensions. The size of a block is fixed and corresponds to theEXTENTSIZE of the table space. This has at least two pages. With a pagesize of 16K, a memory requirement of at least 32K is needed for eachblock.
Multidimensional clustering can improve the performance of databaserequests. When the requests are restricted to MDC dimensions, only theblocks of the table are read that contain the corresponding dimensionvalues. Block indexes are used for data access that are much smallerthan the usual row indexes, which means you can search through them morequickly.
The performance of data manipulations (INSERT, UPDATE, DELETE) alsobenefits from the physical organization of an MDC table, and from theminimal effort required for maintaining the smaller block indexes.
The default secondary indexes are no longer created for the InfoCubedimensions in the ABAP Dictionary. Instead, system-internal blockindexes are generated for these dimensions.
The clustering of the InfoCubes is transferred for all aggregates. If anaggregate does not include an MDC dimension of the InfoCubes or if allInfoObjects of a MDC dimension were created as line item dimensions inthe aggregate, the clustering of the aggregate is created with theremaining MDC dimensions. If an aggregate does not contain MDCdimensions of the InfoCube or if it only contains MDC dimensions, indexclustering is used for the aggregate.

Procedure
Choose a time characteristic for the time dimension or the column withthe dimension IDs (DIMID) for the time dimension. Only 0CALMONTH or0FISCPER can be used as time dimensions. The time characteristic mustexist in the InfoCube.
The time dimension always gets the status number 1.
Choose more dimensions for multidimensional clustering, by assigningsequential status numbers for the dimensions.
Proceed as follows for the selection:

• Choose dimensions by frequently restricting your queries:


• Choose dimensions with a very low cardinality:

The MDC dimension is created in the column with the dimension IDs(DIMID). The cardinality is determined by the number of differentcombinations in the characteristics of the dimension . Thus choose adimension with one or only a few characteristics and with only a fewdifferent characteristic values.
Line item dimensions are generally not suitable because they usuallyhave a characteristic with high cardinality.
If you want to create an MDC dimension specifically on a characteristicwith very low cardinality, you can define the characteristic as a lineitem dimension in the InfoCube. This does deviate from the general rulethat line item dimensions contain characteristics with very highcardinality, but has the advantage specifically for multidimensionalclustering that the fact table contains the SID values for thischaracteristic instead of dimension IDs and that database queries can bedirectly restricted to this SID column.

• You cannot select more than three dimensions including the time
dimension.

• Assign the status numbers according to the following criteria:

Sort the dimension according to frequency of its occurrence in BIqueries, that means assign the lower status number to the InfoObjectthat occurs more frequently in BI requests.
Sort the dimensions according to selectivity, that means, assign thelower status number to the dimension with the higher selectivity.
Note that for every value combination in the MDC dimensions, at leastone block of at least 32K of memory area is created. This memory areais allocated independent of the number of data records that have thesame value combination in the MDC dimensions. If the number of datarecords with the same value combination is not sufficient for fillingthe block completely, the free memory is not used because data recordswith other value combinations in the MDC dimensions cannot be written inthe block.
If there are only a few data records for each combination in theselected MDC dimensions and thus the most of the blocks contain unused,free memory, too much unnecessary memory is used for the fact table.This decreases the performance of database queries because many pageswith not much content have to be read.