PS: My PC is copying about 80GB of data currently, at the unbelievably high data rate of approx. 0.6MByte/s, so it will be finished by, well, Tuesday morning or something like that. Sheesh, I knew Linux raid5 is slow on write operations, but _that_ slow?


Wow, that does sound slow, although I have no idea what numbers you should be getting..

But on my linux raid1 system, on a slow celery 500, a 2GB file just took 2.5 minutes to create, ie ~14MB/second, using approx 30% CPU. I know that calculating parity takes time, but 23 times longer???!

Some pointers - You really need the RAIDed drives all to be on independant IDE channels to be effective. I'm guessing that you already know this, and have additional channels on your mb / add-in card.
Have you checked your HDs DMA settings? Some linux distros don't optimise the HD settings by default, and leave things like DMA, write caching and multiple sector I/O disabled on boot. Use hdparm to find out what your current settings are, and to change them if necessary. Once you have found the optimum settings, I suggest scripting them in /etc/rc/rc.hdparm and making sure that script gets run from rc.sysinit before the fsck would kick in. (fsck'ing a large RAID partition without drive optimisations is painful, in fact if you can use ext3 or reiserfs, then I'd recommend it )

man hdparm is very informative.


Command being timed: "dd if=/dev/zero of=./zerofile bs=4k count=524287"
User time (seconds): 0.80
System time (seconds): 46.54
Percent of CPU this job got: 30%
Elapsed (wall clock) time (h:mm:ss or m:ss): 2:36.20
Average shared text size (kbytes): 0