Captain Obvious: use the same type of drives in your RAID.
I’m performing a dd on all my drives to clear them out for an attempt at using my motherboard’s fakeRaid (AMD). I have 4x500 GB Sata II Disks. The oldest disk was manufactured in 2008, the rest purchased at random, all have a 3 Gbps link and are on ‘separate’ channels.
Nothing fancy, just a standard one pass drive fill with zeros, 1M at a time.
Ex. dd if=/dev/zero of=/dev/sdX bs=1M
sda - Hitachi HDP72505 (The noisy one)
sdb - Western Digital (Caviar Green) WD5000AADS (The eco conscious one)
sdc - Seagate ST3500641NS (The old one)
sdd - Western Digital (Caviar Blue) WD5000AAKS (The new one)
What’s interesting is just how large of a performance gap there is between the drives, with the Hitachi and WD Caviar Green drives having a similar write speed (but lower utilization on the Hitachi), the Seagate is clearly at the bottom of the pack, and the WD Caviar Blue leading. S.M.A.R.T. says all the drives are healthy.
Device: rrqm/s wrqm/s r/s w/s rMB/s wMB/s avgrq-sz avgqu-sz await svctm %util
sda 0.00 176996.00 0.00 1140.50 0.00 83.90 150.66 26.98 23.29 0.67 76.80
sdb 0.00 19211.50 0.00 164.50 0.00 82.05 1021.54 54.74 344.27 6.09 100.15
sdc 0.00 14354.00 0.00 114.00 0.00 55.78 1002.00 41.93 374.39 8.79 100.15
sdd 0.00 28432.50 0.00 227.50 0.00 105.75 951.96 40.46 170.22 4.40 100.20
I also noticed that CPU use was the highest on the wipe of the Hitachi (21%), next highest on the Seagate (9%) then relatively low on both of the WD drives (5-6%).
This is just another example of why you should try to use all the same drives in a RAID. If this were a server of course I’d be heading in that direction, but this is just my desktop.
This isn’t a perfect test by any means, but I thought I’d share…
Update:
sda - 500106780160 bytes (500 GB) copied, 6277.51 s, 79.7 MB/s
sdb - 500107862016 bytes (500 GB) copied, 6815.34 s, 73.4 MB/s
sdc - 500107862016 bytes (500 GB) copied, 10403.3 s, 48.1 MB/s :(
sdd - 500107862016 bytes (500 GB) copied, 5148.57 s, 97.1 MB/s
