Rolling Up Data in BI Accelerator Indexes
As with relational aggregates, the data consistency of the InfoCube and BI accelerator index is based on request handling in the BI system. When you load new data packages (requests) into the InfoCube, these are not immediately available for use in a BI accelerator index for reporting purposes. As with aggregates, the process that writes new data to the BI accelerator index is rollup.
If you replace the relational aggregates in an InfoCube with a BI accelerator index, you do not have to make further changes in the process chains or other settings. The process and the associated programs are identical.
The compression of data packages after rollup, as performed with aggregates to improve efficiency does not apply to BI accelerator indexes because the data on the BI accelerator server already exists in a read-optimized format. However, it is useful to rebuild the BI accelerator index if the InfoCube is compressed heavily after rollup
You can use delta indexes to speed up the rollup process.
- New data packages (requests) have been loaded into an InfoCube.
- BI accelerator indexes for this InfoCube have been activated and filled with data.
When you rollup data for an InfoCube, the system first loads the new data into any aggregates that exist in the InfoCube, and then determines the delta of the missing records for all the tables that have an index in the BI accelerator index of the InfoCube and indexes it. If new SIDs are generated when transaction data is loaded, the system also writes new records to the indexes of the S, X and Y tables. When the system has indexed all the indexes successfully, the data of the most recent request is released for reporting.
As with relational aggregates, you only have to exit data rollup after loading transaction data.
In InfoCube administration, where you can see whether a rollup is missing, running or successful, the system does not differentiate between whether the InfoCube has aggregates or a BI accelerator index.