Hi

We have a rather complicated cleaning process. We have implemented it so that we divide key space to 4 separate number spaces and run them concurrently. Each of the 4 processes make decision if that row is to be deleted and if so deletes the row and dependent rows by cascade delete. Because there were occasional deadlocks each deletion is in its own transaction. Code is PL/SQL, so no waits for writing redo logs should be happening (commit point optimization).

However in some point of time replication of database was changed from materialized view to Golden Gate (volumes got much bigger and materialized views had serious difficulties in handling them).

Now we have encountered issues with Golden Gate, lags of replication when large deletes are done.

My question is: if in source system commit point optimization is used, it is also applied in target system ? Should we try to increase interval between commits so number of waits in redo logs writes in target system would get smaller ? Something else ?


lh