I have been experiencing some disk performance issues with one of the linux servers I've inherited. Having done some simple analysis, I believe this to be because the disks are incorrectly recognised as ide instead of sata. I.E., they are /dev/hd* and not /dev/sd*.

using hdparm i get the following timings..

/dev/hda:
Timing cached reads: 5212 MB in 2.00 seconds = 2607.91 MB/sec
Timing buffered disk reads: 12 MB in 3.49 seconds = 3.44 MB/sec

/dev/hdc:
Timing cached reads: 5192 MB in 2.00 seconds = 2596.55 MB/sec
Timing buffered disk reads: 12 MB in 3.36 seconds = 3.57 MB/sec


im fairly sure i can improve this by instructing fedora to use a different drivers which would then have my disks detected as sda and sdc respectively.

my problem here is that the disks in question are part of a software raid configuration...

Personalities : [raid1] [raid6] [raid5] [raid4]
md0 : active raid1 hdc2[0] hda2[1]
4192896 blocks [2/2] [UU]

md1 : active raid1 hdc1[0] hda1[1]
239994880 blocks [2/2] [UU]


If i were to change the drivers, how do i reconfigure the raid to use sd* instead of hd*?

Hey!
Yep, 4 MB/s makes no sense.
I think it might be that the system would see the drives and automagically understand from the header on the HDDs that they belong to a raid and automatically start the raid
or
it might be that somewhere under /etc/ the mdadm configuration file lists their names (or their UUID) and the raid-device (e.g. md0) they belong to
or
you have somewhere a custom script which builds the raid.

ABSOLUTELY TAKE A BACKUP OF THE DATA STORED ON THE RAID BEFORE FIDDLING AROUND WITH IT!!!
This because I'm not sure if changing the devices from HD* to SD* will screw up their UUID which might be used to recognize the components of the raid.

And prepare youself to use mdadm or whichever raid-tool you're using to fix all problems that might arise.