Hello
I am afraid there is no simple answer.
The CDC refresh is not particularly fast, particularly when compared with dedicated mass data movement tools such as Aspera or IBM DataStage, partly because there is no inherent parallelisation, and partly because unlike snapshot only replication, the refresh often needs to be done while active (concurrent DML operations on the source table) and this is an additional complication that needs to be handed properly.
The native CDC refresh will remember where in the transaction log a refresh of a table started and finished and when mirroring is started, any changes for the table occurring between the start and finish refresh points (in doubt transactions) will be applied to the target database adaptively if possible (upsert logic). The LUW agents support the adaptive apply of in doubnt transactions during an external unload/load process by providing commands to mark the start and end of the unload of the source data.
Sometimes the CDC refresh is the only option, because there is complex transformation or filtering to be done that cannot easily be reproduced by an external unload/load process. In that case there are somewhat complicated ways of introducing parallelism by having multiple concurrent refresh subscriptions. The tables are mirrored all together in one normal subscription. In order to deal with concurrent changes, the parallel refresh is treated as an external unload/load action.
So it all depends...
Regards
------------------------------
Robert Philo
------------------------------
Original Message:
Sent: Thu May 27, 2021 07:26 AM
From: manish bachhania
Subject: CDC Replication for Cloud Target
Hello Team,
We have a requirement to move the data from on-premises data sources to a cloud database.
Could you please help to understand, does IIDR is the suggested tool to move the initial data load(in TB's) from on-premise to cloud target?
Thanks,
Manish
------------------------------
manish bachhania
------------------------------
#DataIntegration
#DataReplication