Hi there,
I've just set up a machine with the following configuration:
Mandrake 7.0-2, Linux Kernel 2.2.14 (stock SMP)
Dual Pentium III (Coppermine) 700 MHz
2048M RAM
Adaptec 39160 Dual Channel Ultra160-Wide SCSI Card
4 x 18G Seagate Barracuda Ultra160-Wide SCSI Disks
5-disk SCA backplane connected to Channel A
The disks all have the following partition table:
partition 1 20M /boot
partition 2 4G /
partition 3 14G /home
I'm planning to set the root partition up with RAID-1 across sda2, sdb2,
sdc2 and keep a time-lapsed 'safe' copy on sdd2.
I'd like the /home partition to be as fast as possible, so I thought that
RAID 0+1 was a good solution (giving approximately quadruple read and
double write performance). After experimenting, I got the following
results:
Read (MB/s) Write (MB/s)
1 disk 26.3 24.7
RAID-1, 2 disks 26.3 24.8
RAID-1, 3 disks 26.3 24.3
RAID-0, 2 disks 51.9 51.8
RAID-0, 3 disks 77.7 69.4
RAID-0, 4 disks 95.6 84.0
RAID-5, 3 disks 51.8 30.5
RAID-5, 4 disks 77.0 31.3
(I never actually benchmarked RAID 0+1 because I assumed it was pretty
obvious based on the above benchmarks)
BTW, I was pleased to discover that Linux had absolutely no problems with
Ultra160-SCSI - 96MB/s isn't bad at all, is it?
I was hoping that RAID-1 would 'stripe' reads between the disks,
increasing read performance to RAID-0 levels, but leaving write
performance at single-disk levels. Does anyone know why it doesn't do
this?
If RAID-1 doesn't 'stripe' the reads, then it seems that my best bet is
going for RAID-5 over 4 disks. Does anyone have any comments and/or
suggestions about doing this?
Thanks in advance,
Corin
/------------------------+-------------------------------------\
| Corin Hartland-Swann | Mobile: +44 (0) 79 5854 0027 |
| Commerce Internet Ltd | Tel: +44 (0) 20 7491 2000 |
| 22 Cavendish Buildings | Fax: +44 (0) 20 7491 2010 |
| Gilbert Street | |
| Mayfair | Web: http://www.commerce.uk.net/ |
| London W1Y 1FE | E-Mail: [EMAIL PROTECTED] |
\------------------------+-------------------------------------/
