Drive recomendations

[tbrezniak]

I know people always argue about this topic. But is it really worth the 20-30 premium per drive to purchase WD RED drive instead of purchasing WD Gren Drives?

 

[L]

depends on what you are doing ..

 

[tbrezniak]

I will be making a 4-5 drive raid 5 or raid 6 array for my movies and plex

 

[L]

To me that's what the green's were designed for, BUT... zfs might freak out when the drive isn't there, or parked. You might see errors when it thinks the drive is gone rather than parked, might see some resilver starts.

 

[Ericloewe]

It boils down to this, according to general opinion:

Would you like an extra year of warranty for 30 bucks? Do you need TLER? Would you like a chance of not having to mess with the drives' head park timers?

If you answered yes on any of the above, go with Red. Otherwise, go with whatever your heart says (man, that's a cheesy line).

For what it's worth, Cyberjock says his rather large collection of WD Greens has not been problematic with FreeNAS and ZFS, despite not having TLER - which leads me to believe it's not really a necessity. I can imagine it would have an impact where availability and performance consistency are important.

 

[Z300M]

I know people always argue about this topic. But is it really worth the 20-30 premium per drive to purchase WD RED drive instead of purchasing WD Gren Drives?

I don't know about other capacities, but for 4TB NAS drives, the HGST (Hitachi) are getting far better ratings on NewEgg than the WD Reds: all 4-egg and 5-egg ratings for HGST, 28% 1-egg (including many DOA) for WD Red.

 

[cyberjock]

I don't know about other capacities, but for 4TB NAS drives, the HGST (Hitachi) are getting far better ratings on NewEgg than the WD Reds: all 4-egg and 5-egg ratings for HGST, 28% 1-egg (including many DOA) for WD Red.

Then go with the Hitachi drives. Trying to argue over hard drive brands is like arguing over religion. At the end of the day, everyone will always think their religion is the best and the rest suck. So avoid the war and just go with what makes you happy. It's your data, for better or for worse.

I hate Seagates with a passion. A serious passion. But if that's what you want to buy, go for it. I just know that I personally wouldn't trust my data with Seagate. Likewise I've seen several people in this forum admonish WD. But I love WD. My old setup had 24 of them and I had amazing reliability with them. Needless to say I bought 10 WD Reds without question just 3 months ago. :P

 

[Chris McLeod]

Just a little food for thought (and hopefully not to fan the flames of religious fury) check the link for a reliability report by Backblaze, the cloud backup provider. They use a LOT of HDD's from various manufacturers and the data is pretty interesting.

https://www.backblaze.com/blog/hard-drive-reliability-update-september-2014/

 

[Ericloewe]

Just a little food for thought (and hopefully not to fan the flames of religious fury) check the link for a reliability report by Backblaze, the cloud backup provider. They use a LOT of HDD's from various manufacturers and the data is pretty interesting.

https://www.backblaze.com/blog/hard-drive-reliability-update-september-2014/

That's another topic that almost sounds like religion over here

 

[ewhac]

Just a little food for thought (and hopefully not to fan the flames of religious fury) check the link for a reliability report by Backblaze, the cloud backup provider.

I don't know why Backblaze keeps getting brought up as the source for drive reliability stats, when Google's farm is at least an order of magnitude larger, and their paper came out years beforehand:

http://static.googleusercontent.com/media/research.google.com/en/us/archive/disk_failures.pdf

It would be nice to see them publish an update, though, particularly one incorporating SSD stats.

 

[Z300M]

I don't know why Backblaze keeps getting brought up as the source for drive reliability stats, when Google's farm is at least an order of magnitude larger, and their paper came out years beforehand:

http://static.googleusercontent.com/media/research.google.com/en/us/archive/disk_failures.pdf

It would be nice to see them publish an update, though, particularly one incorporating SSD stats.

Backblaze is reporting their experience with current- (or at least recent-) model drives. Just because BrandX used to be very reliable a few years ago doesn't mean that it is the best now: higher-capacity drives may be less reliable, or changes in design for drives of the same capacity may lead to reduced reliability. The Google document doesn't help me to decide between BrandX and BrandZ now. Even an updated document of that same general type would not help me.

 

[cyberjock]

I find Backblaze's documents to be almost useless. They do weird things (have a custom build) where they knowingly (and admit it) do things that "aren't recommended" and could be argued that their data doesn't provide anything useful for users looking to do things "properly". Their first public discussion about their hard drive reliability was ripped to utter shreds. Backblaze should have been extremely embarrassed to be such a large user of hard drives and storage yet publish such crap. I could provide better documentation by wiping my butt with some 8.5 x 11 paper.

Google's documentation doesn't talk about specific models at all (it was expressly mentioned early in the document that they would not discuss those aspects). So the concern you have @Z300M with the Google, even if it were up-to-date, would pride nothing for what models or brands to buy. What Google's white paper *does* do is provide information on things that clearly have shown to shorten drive life and things that seem to be inconsequential. For example, the single biggest piece of information I find of value from the Google White Paper is the temperature charts. Until then nobody really paid much attention to hard drive temps. As long as they weren't at the upper end of the designed temperature range then people just figured it was fine. I used to have to replace at least one hard drive a year from one of my 4 machines (each had only 1 hard drive, so 4 disks total in my sample). This closely matches the 15-25% failure rate that people are often dictated with regards to annual failure rates (AFR). But after taking temperatures seriously and keeping drives below the 40C I had 24 drives in my server for 4-5 years. Some were just over 4 and some were just over 5. Despite nearly 24x7 uptime I had a whole 3 failures. That's less than 3% per year for an AFR.

If someone had told me that temperatures mattered that much 10 years ago I would have dismissed them as an idiot. But, I've since learned the ways of drive longevity, and I am glad I found them. Replacing hard drives isn't cheap.

 

[Ericloewe]

The problem (which Google cleverly chose not to tackle) is that it's dammed hard to tell where hard drive populations end.

Take the whole Green vs Red story. Long discussions were held over how different they were and nothing conclusive came of it.
Now consider new models in an old series, like the 5 and 6 TB WD Reds. They seem to be different from the "old" models, which seemingly haven't changed, yet got new"features" like the larger versions.

Basically, defining meaningful populations of drives could almost be a research paper in its own right - one that would quickly be outdated.

 

[cyberjock]

The fact of the matter is nobody has enough information to accurate depict the stats for buying a WD Red vs WD Green. As Seagate clearly demonstrated in 2009, a firmware version that was pushed out and expected to be minor and uneventful lead to so many failed disks that Seagate had to offer data recovery services, for free. I was one of those boned by their screwup. Wasn't pretty for anyone involved. It was so severe that hard drive recovery companies had "leaked" the fact that there was a major problem before Seagate admitted to it. (One place I'd love to be an employee of is a data recovery company. They do cool stuff and I'd love just to watch them do it once.)

So even now, when it comes to buying a Red vs Green, you're literally basing your decision on data from no more recent than a month ago. Look at how long it took to find out WD Reds had the buggy firmware where they had a high load cycle count despite the fact they aren't supposed to do that.

Anyone, quite literally, is dropping big cash for hard drives that, in the big picture, you can't even truly validate are reliable or not until the money is spent and you can't send them back. Real crappy position to be in as a consumer.

Throughout my life I've always had important data that I didn't want to lose due to errors and such. I've lost quite a bit despite trying to do things properly. I'm in my mid 30s and twice lost massive amounts of irreplaceable data. In the first case it was out of my control as the data was lost *because* I wanted a backup. In the second case it was because I was serving my country and while gone I had multiple drive failures in very short succession without the ability to even know something was wrong until I got home and my g/f told me that nothing was working. Talk about coming home after not being home for 3 months to find that the entire server is down.

Thank goodness we have ZFS. For the first time in my life, I feel that my data is truly protected from the most likely outcomes that might cause me to lose my data. I never had great faith in RAID, and I was always using par2s to check the data because I lost faith in RAID controllers years ago. On more than one occasion I had files that were fine when I ran a PAR2 before doing a RAID rebuild, but after rebuild they had corruption. Not a good place to be when those kinds of problems plague hardware RAID.