Is there any way to improve the extraction from AS400 using parallelism with SSIS ETLs?
I'm migrating DTS to ETLs but the performance is so poor... obviously something can be done in order to at least match the performance (in time) of the DTS. But I have no idea how (very new on this)
Any tips please?
Parallelism in SSIS is generally done by writing the package to have multiple operations in the same space (i.e two dataflows with no control flow setting any precedence).
In theory, you could partition your query to do this (split the data in half and have a condition on each flow?). Without knowing anything about the insides of a AS400, I couldn't tell you if this would be worth doing. Potentially you could run extracts from different tables at the same time.
I dont believe that parallel extraction from a single file is possible. If you can use a direct ODBC connection to the AS400 instead, this might give a bigger performance boost over trying to tune a file export/import process.