Disk I/O

[xadamz23]

Hello,

Disk I/O on my linux boxes is very poor. I must not have things configured correctly. I've noticed poor performance on various configurations (EXT3, ReiserFS, XFS) tried with RAID 0, 1, 5. I know that better I/O can be achieved because we have a couple other linux boxes here that were built by other admins that seem to handle I/O just fine.

Here are the specs of one of my boxes:
SLES 10
Dual-Core AMD Opteron 2.6GHz (x2)
8 GB RAM
5 72GB 15K SCSI drives in RAID 5 config

Output of "fdisk -l":
Disk /dev/cciss/c0d0: 364.1 GB, 364172083200 bytes
255 heads, 63 sectors/track, 44274 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

Device Boot Start End Blocks Id System
/dev/cciss/c0d0p1 * 1 26 208813+ 83 Linux
/dev/cciss/c0d0p2 27 1985 15735667+ 83 Linux
/dev/cciss/c0d0p3 1986 3030 8393962+ 82 Linux swap / Solaris
/dev/cciss/c0d0p4 3031 44274 331292430 8e Linux LVM

Output of "df -h":
Filesystem Size Used Avail Use% Mounted on
/dev/cciss/c0d0p2 15G 4.2G 9.9G 30% /
udev 3.9G 108K 3.9G 1% /dev
/dev/cciss/c0d0p1 198M 12M 176M 7% /boot
/dev/mapper/vg01-lvoraclebase
311G 105G 191G 36% /oraclebase

All are EXT3 filesystems.

When I have any large I/O writes that take place, the system becomes unresponsive until the write is done. I know that RAID 0 would be better, but I have experimented with that as well and didnt notice much difference. I have noticed poor disk I/O on many different boxes. We have much older Unix boxes running on much older harder that can outperform our latest and greatest hardware running linux, as far as disk I/O that is. What could I be doing wrong? Is there anything that I can tune in /proc?

[gogalthorp]

Are you using hardware RAD?? Software RAD is known to be faster?

[xadamz23]

I am using hardware raid. I'm pretty sure hardware raid is faster than software raid.

[gogalthorp]

Actully not.

http://milek.blogspot.com/2006/08/hw...d-part-ii.html

http://storagemojo.com/?p=222

http://stoilis.blogspot.com/2005/09/...mise-raid.html

http://scotgate.org/?p=82

Also what file system are you using. JFS may be the best choice for thorughput

[xadamz23]

That is interesting. All of my filesystems on the system shown in my first post are EXT3.

[likwid]

ZFS is sun's revolutionary new FS, not past devel release for Linux AFAIK.

Unless you've got lots of processor cycles to waste, hardware RAID is going to be faster. It's not just "Software RAID is faster."

[xadamz23]

I believe I may have found the problem, but I have some more testing to do first. Currently all my filesystems (/, /boot, /data) reside on the same RAID 5 array. I think that if I move / and /boot to a dedicated disk, the system does not become unresponsive. So the layout will be something like this:

RAID 1 Array ( 36GB (x2) )
/
/boot

RAID 5 Array ( 36GB (x4) )
/data
swap

Can anyone confirm that having / on the same disk as the filesystem with high disk I/O cause performance problems?