That method will work for you, but I think your conclusion that the parallel processing will lead to bad performance might not be correct. I have had some experience with data loads for which we used the parallel processing you refer to and the performance scaled nicely, i.e. there was little overhead from the parallel processing mechanics. Of course this will only work if you can separate your work in distinct parts that do not need to redo any work which I assume the 200K records are. Advantage of this approach is that you can control the amount of work processes used through some basis settings and tweak it to your hardware resources without having to change to the program.
↧