Another thing about buying more expensive drives because they have a longer warranty is that to actually take advantage of it you have to ship your drive with its precious confidential data to strangers who have access to high-tech equipment to restore and suck that data off.
How do they dispose of the drives? Melt them? Demagnetize them forensically and securely? Hire an army of unemployed olympic weightlifters with sledgehammers?
RAID is local but I can't do offsite because I need to backup tens of GBs at a time (during a typical photo trip I generate 50-100GB of images) so it's not a small incremental backup every time and it becomes impractical given network speed plus I need to pay for offsite storage. Before the current prices it made a lot of sense but now it's just a pain to pay $200+ per a drive.
speedmaster20d wrote:
How long does it take to rebuild a 3 Disk RAID5 array? I am using a dedicated adaptech RAID controller.
I've never used a RAID5 setup that's comparable to yours, so I don't have any firsthand experience. I would guess that a "good" system (fast disks, capable controller) could rebuild a 2TB disk in a few hours, but I've heard stories of "bad" systems taking a day or more to rebuild.
aborr wrote:
I've never used a RAID5 setup that's comparable to yours, so I don't have any firsthand experience. I would guess that a "good" system (fast disks, capable controller) could rebuild a 2TB disk in a few hours, but I've heard stories of "bad" systems taking days to rebuild.
my system is good (Core i7 extreme 24GB memory etc.) but several hours is a pain...I think I'll have to pick 4X cheaper 2TB drives and go for a RAID10 or just buy a single high end 3TB drive and make due with it until the prices come down.... there is no way I want to spend more than a $1000 for 4 enterprise drives and watch prices drop by a factor of 3 when production picks up again!
Feb 23, 2012 at 07:22 PM
Lars Johnsson Offline • • • • • • • Upload & Sell: Off
I have 3 Drobos. In two of them (backups) I only use Caviar Green drives. Nine 2TB Caviar Green drives. They have been working very good. In the other I have better/faster drives so I can do PP from that Drobo.
I use 16 Caviar Green 2TB drives at home. Most of them are 1-1,5 years old. And none have failed or had any problem yet.
Lars Johnsson wrote:
I have 3 Drobos. In two of them (backups) I only use Caviar Green drives. Nine 2TB Caviar Green drives. They have been working very good. In the other I have better/faster drives so I can do PP from that Drobo.
I use 16 Caviar Green 2TB drives at home. Most of them are 1-1,5 years old. And none have failed or had any problem yet.
Thanks, what's the part number for your 2TB drives?
Again, a writer is confusing anecdotes for evidence.
i am aware of the statistal angle. i'm also aware 99% of the people are not running large enough samplings to make it fully viable. way too many variable to make it work on the curve.
You'll be sorry if you don't use enterprise drives and not because of reliability. Especially in WD product (Cavier), consumer drives often re-calibrate DURING a transfer...in a RAID array, they'll drop out and the array will have to be rebuilt (think 24+ hours). If you want to save a little money and not buy enterprise drives, look into AV drives (they, also, don't re-cal during transfers). BTW, consider going RAID 6 if your controller supports it.
Also, consider that RAID 0 and RAID 1 aren't as fussy (when reviewing the others' experience about Caviers working). My experience is based on RAID 5 arrays in mission-critical applications (destroying $400k of equipment and logging data on the experience).
You might still be able to find 1-1.5TB Samsung (now owned by Seagate) enterprise drives in the supply channel. I've had a couple of HD103SJ's in a RAID 0 array for about 18 months w/o problems and that model IS an enterprise drive so they would be suitable for RAID 5 or 6. Newegg has 'em for around $160 each (considerably more than the $60 I paid pre-tsunami&quake, but about $100 cheaper than anything else in its class).
Feb 24, 2012 at 07:13 AM
Lars Johnsson Offline • • • • • • • Upload & Sell: Off
Check the WD website or ask them a question...they do NOT recommend using Caviar drives in a RAID array.
As for drive prices coming down, remember that Samsung's drive division is now owned by Seagate and Hitachi's drive division is now owned by Western Digital. That leaves two companies standing. My guess is that they'll try to milk it for what its worth until the SSD sector becomes cost competitive enough to drive them out of business.
Also, regarding "anecdotal evidence":
I've never had a Seagate drive fail (I have an ST238 that still works...that is around 20+ years old). I've never had a Hitachi drive fail. I've never had a WD drive "play nicely" with another WD drive (I've never used them in a RAID array), but they seem fine with other brands. My longest running drive (still in use) that has been in use 24/7 for over 8 years now is a Maxtor and that is NOT representative of my experience with them (they had the highest failure rate...another brand consumed by Seagate). So, I don't think there is a clear "winner" out there.
Thanks everyone, I was able to get some 3TB Seagate Barracuda drives for $200 thanks to a newegg promotion. They are not the the Barracuda ES (enterprise) version but they are 7.2 KRPM that support NCQ with 3 year warranty and longer MTBF. I have used this type of drive in a RAID0 config and they were fine...so we'll see. I'm also retiring my 10KRPM drives for a SSD!
good discussion, as usual experiences greatly vary in these things
I have 4 Hitachi HDS722020ALA330 drives in a QNAP TS-419P+ and they have been flawless for over a year with 100% uptime. If you follow the server fora, you will quickly find that Seagate and WD green drives are NOT suitable for raid and QNAP had even taken most of them off their compatibility lists. If I were to set up another system right now, I'd choose the Seagate 3TB model listed here:
irish-george wrote:
Check the WD website or ask them a question...they do NOT recommend using Caviar drives in a RAID array.
As for drive prices coming down, remember that Samsung's drive division is now owned by Seagate and Hitachi's drive division is now owned by Western Digital. That leaves two companies standing. My guess is that they'll try to milk it for what its worth until the SSD sector becomes cost competitive enough to drive them out of business.
Also, regarding "anecdotal evidence":
I've never had a Seagate drive fail (I have an ST238 that still works...that is around 20+ years old). I've never had a Hitachi drive fail. I've never had a WD drive "play nicely" with another WD drive (I've never used them in a RAID array), but they seem fine with other brands. My longest running drive (still in use) that has been in use 24/7 for over 8 years now is a Maxtor and that is NOT representative of my experience with them (they had the highest failure rate...another brand consumed by Seagate). So, I don't think there is a clear "winner" out there....Show more →
guess i missed that one. i'll need to run out now after 2 years and only one recent predicted failure and get a whole batch of new ones. i have 5 working nicely together. but you are correct there is no clear winner. SSD have their own oncoming issues w/o some changes. the ability to create cost effective spinning media has become problematic. the SSD market is being inundated by secondary makers/marketers already. it is a less labor intensive manufacturing regime overall.
i have 8 dead drives on my desk that were bought in the past two years. most of them are Seagates because that is what i bought the most of, purely on price. it's a mix of 1, 1.5, and 2TB drives. i have a WD 6G drive from 1994 that still works. i have a pair of WD 250G drives that still work too. brand matters only a little bit. i have a pile of WD drives inside one large RAID that are now 4yrs old and all work fine. i have a larger pile of Seagate drives across a wide variety of RAID-like devices including Drobos and NAS units that are newer and i see some of the failing. there are plenty of good reasons to use RAID with the most important being that i exceeded 3TB of drive space needed in a single volume a few years ago and my requirements are still growing faster than drives. that's why i have Drobos formatted at 16TB. that puts off my content migration for 3 or 4 years.
Herb...
Monito wrote:
As stated above, expect drives to fail. In the long run, buying less expensive drives like WD Green and accepting a failure or two over the long run is cheaper than buying expensive drives that have long warranties. If a drive lasts one year, it is likely to last five years. In five years, drives will be available with a lot more capacity anyway.
Beware consumer drives in RAID configuration. In general they are setup in such a way that if they have a missed read on a file they will continue trying to read the drive sector many times before they give up. When you think about it this is a logical way to run a desktop PC drive, as you want it to really try hard before it declares your data as "missing". And in most cases it will actually read the data eventually.
The problem is that when the drive does this it becomes unresponsive to additional requests from the disk controller. This causes RAID controllers (both hardware and software) to declare the disk as failed (the usual timeout for this to occur is 4-8 secs i think) and then drop it out of the array. The enterprise (WD ES etc) drives differ from the desktop ones in that they are programmed in firmware to *not* repeatedly retry to recover lost data, as in an RAID configuration the RAID deals with this intrinsically. Accordingly those drives do not get kicked off the array.
For the "Green" versions of desktop drives the problem can be really bad as they tend to spin down and park their heads after a very short time of inactivity (the WD greens are especially bad for this). The spin-up time can sometimes exceed the RAID time-out interval, which means in effect that they are kicked off the array for saving power. Also, repeated spin-downs and spin-ups causes one of the SMART parameters (Load Cycle Count) to become very large. This can flag up as a SMART error if it gets too big, also causing problems.
Whilst it is only one drive that gets booted off the array this is just an inconvenience that lasts the time of the rebuild (rebuilds max out at around the read-write speeds of the drives, so this can take several hours on a big array). However, if a second drive has a hiccup whilst the RAID array is rebuilding from the first you stand a good chance of losing all your data.
It was possible for earlier models of the WD greens to flash the relevant timeout into the firmware and disable them from spinning down so aggressively. WD even made tools to do this. In more recent models you can't do it though. There is the possibility to program it via SMART tools, but it usually doesn't survive power cycling or reboots. Thus it has to be re-programmed every reboot. Most drives support this, but WD decided at some point to totally de-activate it from their consumer drives (which technically contravenes the SATA spec).
15Bit wrote:
Beware consumer drives in RAID configuration. In general they are setup in such a way that if they have a missed read on a file they will continue trying to read the drive sector many times before they give up. When you think about it this is a logical way to run a desktop PC drive, as you want it to really try hard before it declares your data as "missing". And in most cases it will actually read the data eventually.
The problem is that when the drive does this it becomes unresponsive to additional requests from the disk controller. This causes RAID controllers (both hardware and software) to declare the disk as failed (the usual timeout for this to occur is 4-8 secs i think) and then drop it out of the array. The enterprise (WD ES etc) drives differ from the desktop ones in that they are programmed in firmware to *not* repeatedly retry to recover lost data, as in an RAID configuration the RAID deals with this intrinsically. Accordingly those drives do not get kicked off the array.
For the "Green" versions of desktop drives the problem can be really bad as they tend to spin down and park their heads after a very short time of inactivity (the WD greens are especially bad for this). The spin-up time can sometimes exceed the RAID time-out interval, which means in effect that they are kicked off the array for saving power. Also, repeated spin-downs and spin-ups causes one of the SMART parameters (Load Cycle Count) to become very large. This can flag up as a SMART error if it gets too big, also causing problems.
Whilst it is only one drive that gets booted off the array this is just an inconvenience that lasts the time of the rebuild (rebuilds max out at around the read-write speeds of the drives, so this can take several hours on a big array). However, if a second drive has a hiccup whilst the RAID array is rebuilding from the first you stand a good chance of losing all your data.
It was possible for earlier models of the WD greens to flash the relevant timeout into the firmware and disable them from spinning down so aggressively. WD even made tools to do this. In more recent models you can't do it though. There is the possibility to program it via SMART tools, but it usually doesn't survive power cycling or reboots. Thus it has to be re-programmed every reboot. Most drives support this, but WD decided at some point to totally de-activate it from their consumer drives (which technically contravenes the SATA spec).
I don't belive it works like this at all if you use them in a Drobo (with it's kind of raid) WD Greens work very good in Drobo's. You can also set your spin down time with the Drobo
