Loading Data into Aggregates Efficiently

Setting Automatic Compression

For each InfoCube, you can set whether the aggregates of the InfoCube are compressed automatically when it is filled with data or after the roll up of data packages (requests).

You are in the Data Warehousing Workbench in the Modeling functional area. In the InfoProvider tree, navigate to the required InfoCube.
In the context menu of the InfoCube, choose Manage.
In the lower part of the screen, select the Roll Up tab page.
Under the Aggregates group header, set the corresponding indicator in the Compress After Roll Up field.

Note

Alternatively you can set automatic compression after roll up. This is described below:

You are in the Data Warehousing Workbench in the Modeling area. In the context menu of the required InfoCube, choose Display or Change. Choose Environment InfoProvider Properties Display or Change. On the Roll Up tab page, choose option Compress After Roll Up.

Indicator	Information
Set (default): Automatic compression switched on	The aggregates of an InfoCube are compressed automatically when the InfoCube is filled with data or after the roll up of data packages (requests). If you want to delete a data package (request) from the InfoCube and the InfoCube has already been rolled up to the aggregate, you have to deactivate the aggregate and build it again.
Not set: Automatic compression switched off	The aggregates are only compressed with the InfoCube. Use this setting if you have to frequently delete requests from the InfoCube. A specific request can be deleted from the aggregates when it has been deleted from the InfoCube. Note the possible affects on performance; aggregates can become quite large if they are not compressed automatically.

Indicator

Information

Set (default):

Automatic compression switched on

The aggregates of an InfoCube are compressed automatically when the InfoCube is filled with data or after the roll up of data packages (requests).

If you want to delete a data package (request) from the InfoCube and the InfoCube has already been rolled up to the aggregate, you have to deactivate the aggregate and build it again.

Not set:

Automatic compression switched off

The aggregates are only compressed with the InfoCube.

Use this setting if you have to frequently delete requests from the InfoCube. A specific request can be deleted from the aggregates when it has been deleted from the InfoCube.

Note the possible affects on performance; aggregates can become quite large if they are not compressed automatically.

Reading the Data in Blocks

If the amount of data is very large when you fill the InfoCube, the system reads the data in blocks and not all at one time. This avoids problems with temporary table space on the database which may occur if you have very large sources (InfoCubes or aggregates).

For more information about the block size settings, see Customizing under Start of the navigation path SAP Customizing Implementation Guide Next navigation step SAP NetWeaver Business Warehouse Performance Settings Parameters for Aggregates End of the navigation path .

Optimizing Performance of Aggregates with Fewer Characteristics

Aggregates with fewer than 14 characteristics are created for all databases in such a way that each characteristic is in a separate dimension (artificial) and these dimensions are created as line item dimensions. Aggregates that only consist of line item dimensions are filled purely from the database. This improves the performance when filling and rolling up.

The logical tree display in the right part of the Maintenance for Aggregate screen is copied from the left part of the Selection Options for Aggregates screen but does not mirror this special form of storage on the database.

Optimizing Data Load Performance

To optimize data load performance, you can specify that you want to automatically delete indexes before the load operation and recreate them when the data load is complete. Building indexes in this way accelerates the data load process, although it has a negative impact on system performance when the data is read. Only use this method if no read process takes place during the data load.

If you want to switch on index building during roll up anyway, you have the following options:

You are in the Data Warehousing Workbench in the Modeling area. In the context menu of the required InfoCube, choose Display or Change. Choose Environment InfoProvider Properties Display or Change . On the Database Performance tab page, choose the options Delete index before each data load and then recreate or Delete index before each data load and then recreate.
You are in the Data Warehousing Workbench in the Modeling area. In the context menu of the required InfoCube, choose Manage. On the Performance tab page, choose the option Create Index (Batch) and select the required options: Delete InfoCube Indexes Before Each Data Load and then Refresh or Also Delete and then Refresh Indexes with Each Delta Upload.

Parallel Execution of Processes for More Than One Aggregate

In BW Background Management (transaction RSBATCH), you can specify that the following processes are executed in parallel. These processes all serve to process aggregates. Parallel processing is applied to the aggregates in any number of InfoCubes.

Process Type	Description
AGGRFILL	Initial filling of aggregates
ATTRIBCHAN	Attribute change run
CHECKAGGR	Check aggregates during roll up
CONDAGGR	Compress aggregates
ROLLUP	Roll Up

Note

For roll up you can also make these settings in the InfoCube:

You are in the Data Warehousing Workbench in the Modeling area. In the context menu of the required InfoCube, choose Manage. On the Roll Up tab page, choose . A dialog box appears in which you can define settings for parallel processing.

For the change run, you can also make the settings in the Administration functional area of the Data Warehousing Workbench. Go to Change Run: Under the group header Executing Change Runs, choose . A dialog box appears in which you can define settings for parallel processing.

By default, the system executes a maximum of three parallel processes. You can change this setting (Number of Processes) for each individual process type. In process chains, the affected setting can be overridden for each of the processes listed above.

Caution

Note that fill, roll up and change run each consist of several subprocesses, all of which are processed in parallel.

Example

For example, roll up consists of the following subprocesses:

Roll up data into an aggregate
Compress, as required
Check, as required

The parallel processing settings for the subjobs correspond to the parallel processing settings for the main job. For example, if you decide that you want to perform roll up in five parallel processes and compression in two, the system executes the compress subprocess of the roll up in five parallel processes.

Recommendation

If you do not want the system to respond in this way, you can set parameters for the InfoCube so that the system does not automatically compress the aggregate (see section Setting Automatic Compression above). In addition, you can add the Compress Aggregate process as a subsequent process to the Roll Up process in a process chain. In this case, the system applies the compression settings that you set in the BW Background Management transaction (transaction RSBATCH). In the example above, the system executes roll up in five parallel processes and compression in two.

The parallel processes are executed in the background, even if the main process is executed in the dialog. This can considerably decrease execution time for these processes. You can determine the degree of parallelization and specify the server on which the processes are to run and with which priority (job category). Job category 'A' has the highest priority, followed by category 'B' and finally 'C'.

Caution

Note that if you choose more than two parallel processes (Number of Processes), one process monitors the other processes and divides the work packages. You always have one process less in actual usage than the number of processes selected in the settings.