Performance Tips for Processing Master Data

To ensure good performance when processing master data, we recommend that you note the following:

• Entering master data directly:

  If the number of master data records for an InfoObject is very small, you can enter this master data directly in the system without loading it.

• Loading master data before transaction data:

  Before starting to load the transaction data, you should load all related master data to your BW system. First load the attributes, then the texts and then the hierarchies. If no master data has been loaded yet when the transaction data is loaded, the load process can take twice as long, as the SIDs need to be taken for the characteristic attributes, and new records have to be inserted in the master data tables.

• Loading larger amounts of master data:

  First load a smaller amount of master data to the master data tables and rebuild the CBO statistics for these tables. You can then load the larger amount of master data.

• DataStore object before master data:

  There are two cases where it can make sense to load the master data to a DataStore object first and then to the master data tables:

  • If you are modeling characteristics with a very large number of attributes and have no need for all these attributes in the analysis. In this case, it is better to save this information in a DataStore object and only to take over to the reporting layer the InfoObjects that you really need.

  • If you only need certain InfoObjects for reading data, but not for the analysis.

  You can then load the data to the PSA in parallel and load from multiple sources.

• Parallelization:

  You can extract master data from the source system in parallel in order to update data more quickly in the BI system. To do this, create InfoPackages that extract disjunct numbers of master data records on account of the relevant selection criteria, and schedule them in parallel.

  More information: SAP Note 421419

• Master data with a large number of navigation attributes:

  When using complex master data with a large number of navigation attributes, the activation and change run can be very time-consuming. The (complex) SQL statement required for this can be split into a number of less complex SQL statements. You can set this by making an entry in table RSADMIN.

  More information: SAP Note 536223

• Number range buffer:

  If you are loading large amounts of data to an InfoCube, for example for the initial load, you should increase the number range buffer for the InfoObjects in question. This also applies for loading master data. This reduces accesses to the database. After loading these large amounts of data, you should reset the number range buffer.

  More information: SAP Note 130253

• Reorganization:

  Reorganize the master data. The reorganization process finds and removes redundant data records in the attribute tables and text tables. This reduces the volume of data and improves performance.

  More information: Reorganizing Master Data

• Optimizing SID creations for MPP databases:

  In customizing, you can define a threshold value: how many records trigger a mass access when the system reads SIDs. SID tables are usually buffered on a record-by-record basis. If a large number of values needs to be read in the SID table, then single record access (select single) is considerably faster than mass access (select for all entries), provided that all the values to be read are contained in the single record buffer. If the values are not contained in the buffer, mass access is considerably faster than single record access. The relative access times depend very much on the database used.

  In customizing, choose  SAP Customizing Implementation Guide  SAP NetWeaver  Business Warehouse  Performance Settings  Optimize SID Generation for MPP Databases .