If your worried about performace, I would use the load command instead of the import, its allot faster than import, it does it in pages, but on the down side it doesn't do logging so you would need to backup up the tablespace/db when finished to have a ref starting point. It loads the data in by pages.
In my experience, for almost any data-import performance problem, the root cause of the problem isn't the capacity of the hardware, nor particularly its speed, but the sheer volume of work that the computer is obliged to do in order to fulfill the request as-given. If the computer (any computer, any system...) has to update the index structures each and every time, performance is going to be terrible because the disk-drive inevitably thrashes itself to pieces.
(As my mentor once said, "you just run out of milliseconds.")
If, on the other hand, those indexes are turned-off temporarily, the data is loaded, and the indexes are then reactivated, the process is efficient because it only takes one sequential pass through the data to completely rebuild an index.
If the data being presented to the import process is, furthermore, sorted in advance into primary-key order, the operation is faster still. (The sorting is performed by an external sort-tool.) Sorting is an "unexpectedly efficient" operation for a computer to perform, and indexing algorithms can work far more efficiently if the data is in sequence.