FireFoxx74 wrote:Hi all,
Has anyone run Western Digital SE drives in a RAID array? WD warns that the SE drives have a long error-recovery time, which may cause the RAID array to drop the drive, but has anyone actually had that happen?
I am looking at purchasing a Asus M2N32 WS Pro board (6xSATA), and some 500/750/1Tb drives for RAID, but the cost between SE and RE2 is nearly £60 (per drive!).
Any feedback appreciated.
Here is some back story wrote:I run a number of Western Digital Enterprise drives, from the 74GB Raptors up through the 500GB RE2s. All my RAID arrays are based on various 3ware cards.
If you run a 3ware card, the "RE" is absolutely critical. If not an "RE" drive, the drive BIOS must be RAID controller compatible. Many consumer drives will not respond to the 3ware fast enough and the 3ware will drop the drive out. So you will either see a degraded array or no array depending on your RAID setup. This will happen even if you are copying a lot of files, running chkdsk, etc. It doesn't take much. With all the "RE" drives I have used, I have never had a drive drop out due to RAID controller / drive interaction issues.
Other than Western Digital, I've had good luck with Barracuda 7200.10 drives (non-ES actually, many BIOS versions work just find with the 3ware). These Seagate 7200.10s have actually more reliable than my WD 7200 rpm drives, but not as reliable as the Raptors. The "ES" and "ES.2" drives from Seagate also feature time limited error recovery, similar to Western Digital.
The consumer model Hitachi 1TB (7K1000) does not work with 3ware well. But the enterprise model (A7K1000) allegedly does. I didn't order an A7K1000 as the basic drive design runs very hot. It is a fast drive, but for 1TB something like the 3 platter Samsung will probably be the overall winner (which may have issues with RAID as other Samsungs have had in the past).
Outside of the 3ware world, I don't know what will/won't work. You will need to look at how each RAID controller/driver deals with drive access, response time, etc.
At least in the US, Western Digital warranty service on the Enterprise drives is not very fast. It will take 3-4 weeks to get a replacement drive under warranty. You will also have to hassle them once or twice. Hence think "cold spare" as well as "hot spare".
Lastly, I believe on the 3ware website they keep a list of all validated drives. The "SE" drives are on the list. As is a comment "RE is recommended". If a drive works with a 3ware, it will likely work with less demanding RAID controllers/drivers as well.
Okay, one more thing. If you are running RAID, do not buy drives that are brand new to the market. Wait for a few batches so they can get the manufacturing bugs out. This usually takes 6 months (for example, my 7200.10s were purchased after Seagate fixed the early noise/heat issues. The ones I have are really good drives. Fast, quiet, and reliable.) And of course, if you buy drives in quantity make sure to allocate drive serial numbers appropriately amongst your RAID arrays so the probability of coincident failure is minimized.
For your board, the RAID is really host raid. The chance that host raid will drop out with a consumer drive is low. The drivers used are basically the same drivers as for normal non-RAID usage. So the drivers are aware of long time-outs. You should be able to go ahead and get consumer drives for your consumer RAID. Just keep in mind that consumer drives will not necessarily work well on a hardware RAID controller if you ever decide to go that route. And when you get data corruption errors on your consumer RAID due to lack of data scrubbing, well, at least you know the issues when you want to move to better RAID and/or a better file system such as ZFS which can do data scrubbing.
Okay, another one more thing. Beyond working better with your RAID controller, "Enterprise" drives usually have much better bit error rates. Much better can be an order of magnitude.
For instance, the Barracuda 7200.11's BER (bit error rate) is an order of magnitude worse than that of the Barracuda ES.2, which boasts one permanent read error for every quadrillion bits read.
What that means in practical terms is that, with the 7200.11 drives, a read error is statistically certain for every 12TB read. With a big array, this means you will get data corruption on every full array rebuild.
With the more reliable ES.2 drives in place, the error would occur after the drives read 120TB.
So for ANY drives you buy, take a peek at the bit error rates and see how they compare. These error rates are much more important if your RAID controller / file system does NOT do media/data scrubbing.
Cheers