During loading, note the following with respect to the order:
Firstly, load the master data in the following order: First attributes, then texts, then hierarchies.
Load the master data first and then the transaction data. By doing this, you ensure that the SIDs are created before the transaction data is loaded and not while the transaction data is being loaded.
More information: Performance Tips for Processing Master Data
Using partitioning, you can separate the dataset of a PSA table into several smaller, physically independent, and redundancy-free units. This separation can improve performance when updating data from the PSA and deleting from the PSA. In BW Customizing, under(transaction RSCUSTV6), you can define the number of data records after which a new partition will be created. Only data records from a complete request are stored in a partition, even if its size exceeds the defined partition size. The specified value is therefore a threshold value.
Note that requests cannot be physically deleted from the PSA table until all requests in the PSA partition have been deleted.
More information: Deleting Requests from the PSA
Parallelization During Loading
To perform parallelization while processing various requests during data transfer, you have the following options:
Split up large requests using selection criteria in the InfoPackage
For files: Split up large files into smaller files and load them in parallel
To perform parallelization in a request during data transfer and the load process, you have the following options:
For the data transfer, define the degree of parallelization and the parameters for the data package size.
You can find more information about SAP and BW source systems under Performance Tips for Data Transfer from an SAP Source System.
For files, set the maximum number of records in a data package in BW Customizing, under Processing tab in the InfoPackage, where you can define the maximum number of parallel background processes to use for data transfer.(transaction RSCUSTV6). For DataSources (object type R3TR RSDS), select the
For 3.x DataSources, select the Processing tab in the InfoPackage, where you can define parallelization by selectig the processing option. On the Data Targets tab, you can select various targets for a load process.
You should transform data as early as possible in the data flow. If you need certain data transformations from a single source for multiple InfoProviders, you can place an InfoSource inbetween, so that only one transformation is required. You can then make further InfoProvider-specific transformations.
If you use routines in the transformation, you need to ensure that the coding does not impair performance.
More information: Performance Tips for ABAP Programming