Poor SATA-300 500GB RAID1 Performance

ilumos

Distinguished
Sep 10, 2008
1
0
18,510
Hi,

I've recently bought a pair of 500GB Seagate 7200.10 drives (ST3500630AS) to run in a RAID1 mirror. Having set all this up correctly, and bench marked the drives with several different pieces of software (HD Tune/HDTach) the burst and average speeds are far worse than when a single drive is run:

Burst is 170MBps and average read is 80MBps.

When one of the two drives is unplugged (or run outside of the RAID array) the speeds jump up to 245MBps Burst and 90MBps average read.

Why could this be?

I'm running the drives on an Asus A8N-SLI Premium motherboard on the NVRAID controller.

Both drives have their jumpers set to allow for SATA-II (3Gbps/300MBps) operation,
The BIOS is set for SATA-II
I've tried different combinations of SATA ports (four of them) and stripe sizes (to no avail) and the latest chipset drivers for my motherboard.

My previous setup was exactly the same, but with two 250GB Seagate Drives (ST3250620AS) in a RAID1 mirror, and yielded benchmarks that were only 5MBps or so slower than a single drive of the same model. (Burst of 240MBps, average of 80MBps)

If I run the two 500GB drives in a RAID0 stripe I get blistering results (230MBps burst, 160MBps average) but only when the stripe is set up in the NVRAID BIOS... If I set the stripe up in MediaShield (nVidia's software configuration solution for RAID) the speeds are 111MBps burst and 21MBps average, which is slower than ATA-33 ffs.

Sadly the same doesn't happen when I configure the RAID1 mirror in the BIOS, it's just as slow whether I use MediaSheild or not.

If I didn't need the redundancy that RAID1 mirroring offers I'd run the stripe, but it's looking more and more like it's the best option... Despite doubling the chances of data loss.

What could this be caused by? I'm at my wit's end.

Many many thanks in advance,

ilumos
 

chookman

Distinguished
Mar 23, 2007
3,319
0
20,790
RAID-1 writes will write the same data to both drives and im not sure on the reads but i think it actually reads both drives to check the data is intact (someone else may be able to confirm this) so you would expect RAID-1 to be at best equal with a single drive however, more than likely read/writes will be worse than a single drive. So your figures dont suprise me at all. BTW HD Tune/HDTach are only sythetic benchmarks and should be taken with a grain of salt.