Re: Running import datapump over a database link.

> This is the par file am using...

...
> PARALLEL=4
Random thoughts:

As a point of order, you should verify that data pump is actually using 4 workers. From what you've shown us, I'd bet it has created 4 workers, but only 1 is active for the majority of the import. That would certainly increase your runtime. The network throughput should also be monitored during the import to see if 4 concurrent threads could be reasonably used before the bandwidth limit is reached.

If possible, make sure the destination table in your test database has no indexes and ideally no constraints. Recreate those after the import completes (which is an exercise in itself).

Partitioning seems like a possible solution, albeit an expensive one.

As others have said, your use of the MOD function would probably not allow the index to be used. If this is a one-shot import, you could manually create ranges of IDs to be used. If not, those may need to be recalculated every time the import is run.

End of random thoughts...

GL!

Rich

--
http://www.freelists.org/webpage/oracle-l