I has a data generator with a subset and all the rules of the generador are sql. When I execute the generator trought Gt Datamaker, it finish the publish in 2 minutes, but when a execute the same generator but by CA TDM Portal it complete the publish in 28 minutes.
I tried modify the wrapper.java.initmemory and wrapper.java.maxmemory according to:
TDM Portal Performance
Increasing CA TDM Portal JVM memory - CA Knowledge
It is only one iteracction but I tried modify the tdmweb.publish.iterationsBeforeCommit option to but que slow persist.
The subset has 44 tables, and it delay in inserting each record because when more rows has a table more slower is.
Nota: When I executed the same generator by GT Datamaker it is faster and take 2 minutes maximum.
What version of TDM Datamaker and Portal do you have installed?
Where is the repo located? Is it local to the TDM server or remote? SQL or Oracle?
I has the versión 188.8.131.52 of the TDM Portal (This versión is a specific versión that solve a issue with data painter)
And the versión 184.108.40.206 of GT Datamaker
The repo is located in a remote server, both the GT Datamaker and TDM portal are in the same server and it connet to a repository in other server in a SQL database.
Are you using the same connection profiles in both Datamaker and TDM Portal?
I'm using a connection to the same database but the profile was created in both Datamaker and TDM Portal in each one. But it has the same conection configurations.
I am struck by your observation that:
‘The subset has 44 tables, and it delay in inserting each record because when more rows has a table more slower is.'
I have sometimes seen situations where fresh target databases are extremely sensitive to bulk inserts when they are first being populated. Until they reach a certain size, the database statistics have been refreshed and the optimizer can build a stable execution plan, insert behaviour can be very unpredictable.
You may well have already explored this but it leads me to a couple more questions for clarification please: