Re: Methods of high performance disk I/O wrote:

That being said, what I’d really like to understand is why I was literally able to bring my system down by mapping file 1, writing 700MB of data, mapping file 2, writing 700MB of data, mapping file 3, writing 700MB of data, then repeating from file 1. By the time it had gotten finished with say the 4th file, I had to hit the power button on my machine.

The problem is all this mapping. You’ve mapped somethign in the region
of 2.8 gig of “stuff”, for want of a better word. If you don’t do
something like that carefully, you’ll flush just about everything else
pageable back to disk … QED everything runs like a dog. You need to
make sure that the working set / memory footprint does not grab too much
of physical ram.

Incidentally, if you had stopped the file mapping, and left the system,
you would have found that it would eventually get back to a relatlively
happy state. It might have thrashed the disk for about the same amount
of time it takes to boot the machine, but it would have got there …