This is how the performance tab looked: http://screencast.com/t/i771bDcc
Nothing seems wrong there, right? It is not the insert statements that take time.
In the meantime we switched to creating the related records ‘manually’, so we use a different foundset for each table, and import them 1 by one.
We just ‘read’ the relation using solutionModel to determine which fields need to be filled in the foreign table.
This makes the import around 3.5 times faster. Giving the client 4GB heap makes no difference.
@ Nicola: yes, this is smart client, but running on the server itself (it is a one time import). Would running the script in a HC still make a difference opposed to that?
Ruben79:
@ Nicola: yes, this is smart client, but running on the server itself (it is a one time import). Would running the script in a HC still make a difference opposed to that?
No, no difference in latency.
If it’s a one time thing I would use SQL inserts.
ngervasi:
If it’s a one time thing I would use SQL inserts.
Yeah but as I said earlier, we have quite some stored calcs and we would need to recalculate all the records then.
Which would make no difference for the total import time, because it is not the sql inserts that take long, it’s the calculations.
The database load is always under 2% load during import.