Checking Data Consistency in SAP BW Accelerator

This section describes how you can run checks to check the consistency of data in SAP BW Accelerator (BWA) with data on the database.

There are various situations and prerequisites for checking data in the BWA:

• You want to check the data in the BWA regularly or after certain processes have run.

  Note

  You can find out more about analyzing incorrect query data in connection with the BWA in SAP Note 1060387 Analysis with incorrect results in BWA queries.

Comparing the large amounts of data in the BWA server and the database server uses up a considerable amount of system resources. To reduce the system load and thus reduce the runtime while analyzing so much data, we recommend combining checks where possible:

• Check facts

  The fact indexes contain the largest amount of data and therefore use up the greatest amount of resources during the check. Use key figures and aggregation by setting the Only Drilldown option on the Totals in BIA and DB tab page to InfoObject = <Iobj>. Use a characteristic with a small number of attributes like CALYEAR for <Iobj>. If the InfoCube contains a large number of key figures, reduce the number of key figures to reduce the system load on the BWA. If the runtime for the check is too long, you can try to reduce the percentage of data to be checked. This is necessary if a key figure overflow occurs during a test, since the key figure type cannot contain the total of all values (example: key figure type = INT2, but the total of all values is greater than 2 000 000 999).

• Check the completeness of the star schema indexes (dimension index, SID index, master data index)

  These indexes can be very large. We advise against running a regular data comparison test, as this uses up a lot of system resources. Use the partly referential integrity in the star schema instead, to find incorrect or missing data records that link to transaction data in the fact table.

  To do this, run the Totals in the BWA test. Run all joins in the star schema and compare the results as complete aggregation on the fact table. This works like a filter, as the result of the aggregation on the fact table is different to the reference result if data records are incorrect or missing from one of the indexes (in the dimension index, for example).

  As with the other checks, reduce the number of key figures that aggregation is run for if the InfoCube contains a large number of key figures. If the key figure types cannot contain the aggregated values of all data records, and there is an overflow, reduce the percentage of the data to be checked.

• Data consistency in complex situations:

  You can also use the Random Queries check. This uses complex conditions to checks the BWA data. Note that this is a random check. The performance of this check depends on the performance of the query in the database. We recommend using this check if the InfoCube still contains aggregates, as at least one part of the randomly generated queries can be run efficiently on the database. Enter a start value for the random generator. This ensures that other queries can be checked in subsequent checks. The random start value is recorded. This allows the same check to be repeated if the data does not change.