Parallel data apply processing preparations
When replicating changed data capture data, a large proportion of latency is usually attributable to the apply of the changes to the target database.
Especially with cloud systems, the performance is essentially dependent on the throughput of the network connection.
-
Preparation of changed data,
-
Encrypting the data stream, as well as
-
Sending and applying the changed data
significantly contributes to increasing the throughput of the replication.
This option is used for preparing the later parallel processing. It is available for archived log processes and Active Logs/Realtime CDC processes writing data in internal format into RDRS Pipes (see Write Internal Format Data into RDRS Pipe).
The Transaction Management needs to be active.
Tagging of transactions for n parallel apply processes: Each transaction written to the RDRS Pipe is tagged with a number (1-n). The assignment of the numbers is based upon the Round-Robin method.
It should be noted that a parallelization of the maintenance processes can endanger the referential integrity of the data. It is therefore necessary to determine at an organizational level to what extent referential integrity is at risk and finding solutions the integrity can be regained.