AVS Forum banner

WD RED failures rise to 8.8% in new backblaze reliability report, Seagate sucks too.

4K views 38 replies 13 participants last post by  Mfusick 
#1 ·
#2 ·
What is somewhat disturbing (or interesting) is that the average age in years of the WD red's are only .5. This is basically the same as it was last time, with not a lot of time on the clock for these drives. What will the stats look like when it gets to average age of 2 years, like the Seagates are. It would possibly appear that the seagates had some failures starting around the 2 year mark , since the average age is that, and the failure rate has increased from last time.

Hitachi is however consistently good across models, time, and other factors. They have 4706 of the 2TB, and 4593 units of the 3TB, and over 5000 of the 4TB's all in operation so I would give some credit to Hitachi on reliability based on this study. The average failure rate for Hitachi as a brand appears like it's about 1%, that seems pretty darn good to me. Much better than WD or Seagate.
 
#4 ·
What is somewhat disturbing (or interesting) is that the average age in years of the WD red's are only .5. This is basically the same as it was last time, with not a lot of time on the clock for these drives. What will the stats look like when it gets to average age of 2 years, like the Seagates are.
Well, in all fairness, WD specifically recommends not using those drives in enclosures with more than 5 drives. ;) :D


It would possibly appear that the seagates had some failures starting around the 2 year mark , since the average age is that, and the failure rate has increased from last time.
The Seagate drives have been sucktastic from the beginning. The 4TB drives have only been in service there long enough to get out of the infant mortality rate, but there aren't any useful long term numbers. The older drives, (3TB and 1.5TB) had terrible numbers in the last report from Backblaze too. (which, incidentally, echos my experience with ST3000DM001 and ST31500341AS)


Hitachi is however consistently good across models, time, and other factors. They have 4706 of the 2TB, and 4593 units of the 3TB, and over 5000 of the 4TB's all in operation so I would give some credit to Hitachi on reliability based on this study. The average failure rate for Hitachi as a brand appears like it's about 1%, that seems pretty darn good to me. Much better than WD or Seagate.
The Hitachi numbers do look good, but I wouldn't go making any broad generalizations or draw any conclusions about the entire brand, based on a study that included 4 models (out of hundreds)
 
#3 ·
I think the notes on Enterprise drives is even more interesting. Remember these are consumer drives operating in a datacenter environment in which the drives are technically not designed for. And yet the consumer drives are outlasting the enterprise drives that ARE supposed to be designed for that environment.

Despite my recent preference for Toshiba Retail Drives (3TB) in my server, I'm happy to report that all my WD Greens and one Samsung Green have yet to encounter any errors going for 26 months now, operating 24/7 (and a few of those WD Greens were in operation prior to the server being completed). I have yet to have a single drive failure in my server.

Remember that your environment (most likely) is drastically different than the BackBlaze enviroinment.
 
#5 ·
Yes I agree. I read that too and was going to comment on that. They had some good examples of how a more expensive enterprise drive is not worth it. I loved how they assumed a 0% failure rate for the example too with the enterprise drive. lol.

I've had good luck with the Toshiba drives too. Although I have 20 of the Seagate 7200.14's and never done an RMA. Most of mine are about 2 years old now too. I guess I have good luck ?

Greens were terrible for me. But according to backblaze the greens are more reliable than the RED's lol. So there goes the theory of everyone paying more for REDS that's been trendy for a year.

CONSUMER drives > ENTERPRISE DRIVES. I've been saying this and doing it for a long time now, I don't regret a thing.
 
#7 ·
Drive_________________________Model_____________Size___Qty___Age__AFR
Seagate_Desktop_HDD.15________ST4000DM000_______4.0TB__9619__0.6___3.00%
Hitachi_Deskstar_7K2000_______HDS722020ALA330___2.0TB__4706__3.4___1.10%
Hitachi_Deskstar_5K3000_______HDS5C3030ALA630___3.0TB__4593__2.1___0.70%
Seagate_Barracuda_7200.14_____ST3000DM001_______3.0TB__3846__1.9__15.70%
HGST_Megascale_4000.B_________HMS5C4040BLE640___4.0TB__2884__0.2___n/a
Hitachi_Deskstar_5K4000_______HDS5C4040ALE630___4.0TB__2627__1.2___1.20%
Seagate_Barracuda_LP__________ST31500541AS______1.5TB__1699__4.3___9.60%
HGST_Megascale_4000___________HHMS5C4040ALE640__4.0TB__1305__0.1___n/a
Hitachi_Deskstar_7K3000_______HDS723030ALA640___3.0TB__1022__2.6___1.40%
Western_Digital_Red___________WD30EFRX__________3.0TB___776__0.5___8.80%
Western_Digital_Caviar_Green__WD10EADS__________1.0TB___476__4.6___3.80%
Seagate_Barracuda_7200.11_____ST31500341AS______1.5TB___365__4.3__24.90%
Seagate_Barracuda_XT__________ST33000651AS______3.0TB___318__2.2___6.70%


Notes:
Some of these drives were scavenged after the Thailand floods, which probably introduces bias for affected models.
Hitachi Storage was bought by Western Digital and the 3.5" HDD line and some patents were divested to Toshiba for EU monopoly law compliance. HGST no longer exists as a drive manufacturer, but hopefully Toshiba can maintain HGST's stellar reliability record.
 
#8 ·
It's worth noting (and correct me if I am wrong) that the 3TB Toshiba Drives are rebranded Hitachi drives mandated by the WD acquisition of Hitachi. I believe the 2TB Toshiba's are rebranded 2TB Samsungs (?)

Edit: Looks like EricN's data confirms this on the 3TB.
 
#10 ·
The tooling machines and patents were sold from Hitachi to Toshiba as part of the Hitachi/Western Digital merger. It was a monopoly or anti trust measure. TOSHIBA drives are made in the WD plants, part of the sale and agreement included long term lease of that.

So yes you are partially correct that many of the TOSHIBA drives like the 3TB are rebaded Hitachi, or of Hitachi design being manufactured by TOSHIBA in the WD factory. But that would not automatically make them HITACHI AWESOME. Hitachi is pretty well known for quality control and manufacturing prowess, that goes beyond just digital hard drives. The company is known to excel at a MFG level in a lot of other products and industries too. I've had some awesome luck with Hitachi products from routers, to bulldozers. They seem to be a leader in most of the things they do. Back in the RPTV days their CRT for RTPV was considered the best of the best.

I would imagine that 4TB and larger sized drives (like the 5TB Toshiba) are new designs though, since the merger is now a bit old. Toshiba is probably designing new models on their own. They are no stranger to HDD's either. I see TOSHIBA 5TB's on sale for like $150 now all the time. Tempting!

The Hitachi 6TB and 8TB He drives look really awesome, but I need a mortgage to buy them :(
 
#11 ·
Should we switch to enterprise drives?

Assuming we continue to see a failure rate of 15% on these drives, would it make sense to switch to “enterprise” drives instead?

There are two answers to this question:

Today on Amazon, a Seagate 3 TB “enterprise” drive costs $235 versus a Seagate 3 TB “desktop” drive costs $102. Most of the drives we get have a 3-year warranty, making failures a non-issue from a cost perspective for that period. However, even if there were no warranty, a 15% annual failure rate on the consumer “desktop” drive and a 0% failure rate on the “enterprise” drive, the breakeven would be 10 years, which is longer than we expect to even run the drives for.

The assumption that “enterprise” drives would work better than “consumer” drives has not been true in our tests. I analyzed both of these types of drives in our system and found that their failure rates in our environment were very similar — with the “consumer” drives actually being slightly more reliable.
There you have it.
 
#18 ·
It seems like they have consistently poor results compared to average use scenarios, they might want to study more why those storage pods suck ass.

I use the 7200.14, 5900.15, Toshiba 3TB and don't get poor results like that. Perhaps my Norco case is a better environment ???

I'm guessing the anti vibration tech on the hitachi is the primary difference in that set up.
 
#19 ·
Average use for datacenters or average home use?

If you mean for datacenters, I'd like to see the stats you're comparing them with as they seem to be hard to come by.

If you're talking about home use, then... well yeah, of course. Vibration is a huge factor in datacenters. It's not that your norco case is so much better, it's that you don't have it stuffed full of 45 drives. You don't have it crammed in a rack with 9 other cases (with 45 drives apiece) You don't have it supported by a riser, that's probably transferring vibrations from the HVAC system. You don't have it crammed full of high speed fans, because you probably don't want it to sound like a Hoover. From a vibration standpoint, their racks are a nightmare.

Furthermore, while it does appear that the Hitachi drives in the study likely have better resilience to vibration, what good does that do you if you're in an environment with significantly less vibration? It's distinctly possible that the WD or Seagate drives have inferior vibration damping, or compensation, but are superior with regards to temperature or particulate contamination. If that's the case, wouldn't one of those be a better choice for a home storage solution since its an almost certainty that a home solution will be dirtier and warmer than a datacenter? (and also have less vibration)

Also, as mentioned above, their drive appropriation methods could certainly have an impact on the reliability of those drives.

And finally, while I find the Backblaze blog an interesting read if for no other reason than it's nice to see a company so open about it's practices, I find a lot of conclusions that they draw a little dubious. (For instance in their "study" to see if HD temps affected failure rate, they tried to normalize their data to each drives "normal" operating temp which means that if Drive X typically operates 20C warmer than the rest, and it has a higher failure rate, their method minimizes the impact of that model, which doesn't make much sense at all. Plus the vast majority of their drives operate within a 10C window, so if you're trying to see how much of an impact temp has on HDD reliability, I'd think you'd want a greater sample size to actually experience different temps.)
 
#20 ·
Average normal home use. Like we use. Compared to them. Most of what you said is irrelevant (I skimmed it only)

At least for the products they've tested I think it's logical to conclude the Hitachi's are pretty good. In a better environment, I don't see how they would do worse. That part was not making sense to me.
 
#21 ·
Of course it doesn't make any sense to you because you (admittedly) didn't read what I wrote.

I don't find the assertion that a home environment is all around "better" than a datacenter environment. Temps are going to be higher and particulate matter is going to be worse in a typical home environment. I'm not saying that the Hitachis will perform worse in a home environment, but it certainly wouldn't surprise me depending on those drives sensitivity to those particular factors. (which is something we know virtually nothing about based on the Backblaze blog post, which is why I wouldn't draw any concrete conclusions about how any of those drives perform in a home environment based on a their conclusions on their datacenter environment)
 
#22 ·
I guess fundamentally I don't see why the home environment would be worse. My Norco is not hot, it's very well cooled, in a basement and climate controlled home. The room it is located in is never above 72 degrees, ever. Most peoples homes are rather conducive to good environments for hard drives, they are not overly moist/wet, hot, cold, and don't move/shake etc..

I am not making the leap you are why the Hitachi would work potentially worse for someone in a home environment than they do for backblaze. They tested thousands of those too, so the data seems rather conclusive. They sources them over a period of time and from many various sources, just like a consumer might acquire them. The superiority of something like the Hitachi Deskstar 3TB over the WD RED 3TB or the Seagate 3TB seem quite certain, it's statistically a significant difference in reliable. The Hitachi is also older, in use longer, and have actually gotten better over time since last study. The WD and Seagate both got worse in the same time. The fact the RED's are average age only 6 months old too- that is possibly something to worry about too. The Seagates got worse as they approached 2 years in age, so what will the WD RED's do ? WD RED is 8.8% now, which is close to what Seagate was last time.
 
#23 ·
A climate controlled home is not the same as a climate controlled datacenter. I don't know how many datacenters you've been in personally, but I've been in dozens of them and they are almost nothing like a home environment. For starters, they are significantly cooler than any home I've been in. The humidity is precisely controlled, and since there is no one living in the datacenter, there is almost no dust. Nobody is allowed to smoke anywhere near them. Other than the vibrations, a datacenter is an almost perfect environment for hard drives, by design.

People live in houses. People (and their icky dead skin) are the single biggest source of dust in homes. You may not smoke in your home, but plenty of people do. You may have your server located in a perfectly climate controlled basement, but plenty of people have their's stuck in a 80+° closet. So like I said before, outside of the vibration issue, a typical home is a much worse environment than a typical datacenter.

The blog post is great information if you're running a datacenter. But it isn't a scientific study. It has no control group. Their goal isn't even to figure out which drives are the most reliable. They're running a backup business. And the information provided falls into the category of "hey, look what I noticed" You seem to be treating it as if it is some peer-reviewed scientifically controlled study, and on top of that you seem to be willing to extend the authors findings to situations outside of the realm of the data that was taken. Sorry, but I'm not willing to make that same leap of faith that you are without some supporting data, or something to show some connection between the two environments.

What I do know is that corporations spend significant amounts of money to keep their server rooms cooler than a typical home. They spend significant amounts of money to keep their server rooms cleaner than a typical home. And while I have to acknowledge the possibility that Liebert and the rest of Big Air Conditioning have been perpetuating myths about the effects of the environment on hardware longevity, I also haven't seen any data to prove otherwise.

So if you think that everybody keeps their home server in a datacenter, that's fine. Go ahead and believe that.
Or if you think that temperature, humidity, or particulate matter have no effect on the long term reliability of a drive, go head and believe that.
Heck, if you believe that some stats for a few models of hard drives in a very particular environment mean every model from that manufacturer will perform similarly in every environment, go ahead and believe that too.

I happen to know that a typical home server environment is nothing like the typical datacenter environment. And in the absence of any data connecting the two environments with regards to hard drive longevity, I have to acknowledge the possibility that what may hold true in a datacenter, doesn't necessarily hold true in a home.

Either way, I look forward to your response where you once again say you don't understand my point, because you didn't bother to read it.
 
#30 ·
I neither said, nor implied any such thing. I very clearly stated that the Hitachi's could be worse, not that they would be worse. #ThePerilsOfSkimming The point being that we don't know how they would behave in a home server environment, at least not with the level of confidence that we do with regards to those same drives in a datacenter environment. They could be better, they could be the same, they could be worse. We just don't know. Afaik there isn't any similar data on those drives in a home server environment, so any speculation on how they would hold up is purely a guess.

To fall back on the ever popular car analogy, I could take a couple run-of-the-mill sedans and test them on a straight washboard track to see how they fared for passenger comfort. Model A might be far and away the best performer in those conditions but when I take the same two cars and test them on a very curvy track, Model A might very well come in last because the conditions of the first test are very different than the conditions of the second test, and success in the first has no correlation to success in the latter. The drives are no different in that regard.
 
#31 ·
Hmmm... and I just got 2 WD Red 3TB disks: according to the spec sheets, they were the most silent and power efficient ones...

It would be interesting to see the some measurements on vibrations or temperatures in the backblaze installation, just to see what conditions the disks are exposed to. The thing is that they use the disks outside of the recommended environment, and beyond manufacturer recommendations. It seems to me more that this statistic is an indication to how well the disks work outside of recommended conditions.
Over time, I've had failures of WD (ide), Quantum (u160), IBM (u160) and Seagate (u320), but indeed never with Hitachi. Of course, I have far too litle disks to make any statistics, but in general I would say I cannot complain about many failures.
 
#32 ·
It's really tempting to take this information and draw conclusions. As humans we do that all the time based on insufficient data. However, I think the only thing the study really tells us is only how these drives perform in BackBlaze's specific conditions. We might be able to draw conclusions on specific drive models/sizes if we had other large data sets to compare under a variety of other conditions, like large data center, people's home servers in basements, typical desktop workstations, etc., but those data are either proprietary or not available. The drive manufacturers themselves probably have something approaching that kind of information, but they're certainly not going to release it.

The BackBlaze study is interesting to read about, and it's nice to see a company openly discussing its experiences. The only thing I think BackBlaze could tell us that would be directly applicable to us, is how many drives of each make/model were DOA. I don't think they told us that, have they? Once they've installed them in their pods, all bets are off.

I'm going to try to do my best not to be influenced by what they presented when I make decisions on which drives to buy. I think we're stuck considering the same things we had before BackBlaze released their study...total cost, cost per terabyte, length of warranty, and whatever specs are important to each of us. Heck, the ratings on Amazon are probably at least as useful as the BackBlaze study when deciding which drive to buy or not to buy.
 
#34 ·
If your circumstances match that of the test, then yes it makes perfect sense. In the absence of any data that matches your circumstances it is still a reasonable choice to make, as long as you temper your expectations to match that of the data (specifically, since there is no data, then you shouldn't have any expectations as to the reliability of the drives in your environment)
 
#37 ·
My home is climate controlled. Tempt is set to 72 year round. No one smokes. I live in socal so you won't see pollen storms or anything like that going on. It's a pretty stable environment.

Even with it like that, it does not compare to how stable our data center room is which is climate controlled to 64. Humidity is controlled via steam injectors. All air ducts have their filters changed twice a month. If a person walks into that room once a week, that would be a lot of traffic.

Really the only thing my home server setup has that beats my work data center in is noise and vibration, which the data center has plenty of.

To close, I just want to say that I've had 9 WD red 3tb drives running 24/7 for two years now and not a single failure. I have 5 WD 2tb Reds in the same enclosure that has been running even longer. You shouldn't make broad general statements over such a small sample size for such a small amount of time. There are far too many variables here.
 
#38 ·
(Caveat: I didn't read this whole thread) But isn't your situation also suffering from SSS? I can say I had to RMA 2 WD 4TB Red's out of 6 when I first got them. There are always anecdotes both positive and negative. I would put more weight into the backblaze findings only because their sample size is a bit larger, but obviously there are many factors involved.

But part of the reason I also invested in Red's (and the Seagate NAS drives) was for the longer warranties.
 
#39 ·
Warranties do have value but the jump from 2 to 3 years is low value. Usually failures happen under two years or over three years. I'm not sure how often a hdd might fail after 2 but before 3 years to justify a $25+ bump in price when a mfg 5 year replacement for 4 years is $13.99 as extended warranty option on the cheaper drive.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top