Basic object of this document is to explain steps to enable parallelization of data selection of any migration object in TDMS 4.0 on top of DMIS SP09 and above.
By default, the Data Selection activity runs a single work process to select data of a migration object in sender system.
By enabling the parallelization of data selection, you can schedule multiple work processes for access plan calculation, that is, to select the data of a migration object according to:
The number of records in the application table for flat migration objects. (A flat migration object only contains one table.)
The number of records in the TDMS header table for header migration objects.
Steps to Enable Parallelization:
Make an entry for a migration object in the table CNVMBTPRECALCOBJ using the view V_CNVMBTPRECALC.
Specify the following values:
Name of the header table if it is a structured or header migration object. This field must be left blank if it is a flat migration object.
Delimiter value -- maximum number of records that a single process should handle
Name of the migration object
In case "Generation of Runtime Objects" activity is already executed in your package. You need to perform following additional steps too after performing among mentioned steps:
In case data selection job is running for given object, you need to cancel this job from sender system via tcode sm37.
You need to force redefine migration object using troubleshooter available against "Start Data Transfer" activity. it will regenerate objects as per above mentioned configuration setting.
NOTE: There is no need to set P_SEL for any fill header program.
The number of records in the TDMS header table CNVTDMS_05_BKPF is 10 million. (You can find this information in the logs of the corresponding TDMS header activity.)
You want to schedule 5 batch jobs for the data selection of the migration object X_BKPF. In this case, delimiter value should be 2 million and header table name should be CNVTDMS_05_BKPF.
Points to be Noted:
A significant load is placed on the database server of the sender system when multiple processes run for selection on the same table. Thus, number of jobs configured for parallelization of a given migration object should be maximum 20.
Migration object configured for parallelization of selection should NOT have more than 2 tables to achieve better performance results.
Parallelization should NOT be configured for migration object dealing with pool tables.