Backblaze HDD Reliability Stats for Q1 2019 (backblaze.com) 66
AmiMoJo writes: Backblaze's hard drive reliability stats for Q1 2019 are out, and show that Seagate has been improving for some time. It still can't match the long time leader, Hitachi, and had a nasty blip with 4TB drives. The Annualized Failure Rate (AFR) for all the hard drives tested in Q1 was 1.56%. That's as high as the quarterly rate has been since Q4 2017 and its part of an overall upward trend we've seen in the quarterly failure rates over the last few quarters..
Finally! (Score:3, Informative)
News for nerds!
No arguing with the editor's pick on this one.
Re: Finally! (Score:1)
Need 2019 reliability stats for tape recorders and punch card readers.
Re: (Score:2)
I know you can make a recorder out of bamboo, or PVC, but are you sure you can make one out of tape?! Wouldn't that just be a poorly made kazoo?
And if you punch the card reader, you're probably going to go to jail. I don't give a darn how bad your fortune was.
Re: (Score:2)
Traditional HDD's are still the best price per gb/tb for storing backups and near-line data.
Wrong. Its called tape. I'd call you a liar for acting like you know shit when you dont.... but people around here apparently dont like harmful people like you being called out
Re:SSDs (Score:4, Insightful)
Tape, as often as not, is a backup medium that is theoretically better in price per TB, but proven otherwise when one tries to restore the data and it fails due to errors.
Also, tape is not even close to comparable except in an enterprise environment where the cost of the tape device (multiple thousands) can be spread across a large number of tapes. The cross-over point between overall tape storage and cold hard drive storage is >100 TB.
Re: (Score:2)
Tape... the backup medium that used to be much bigger than the average hard drive, but has been playing catch-up since the 90's.
Re: (Score:2)
Tape is a great example of why tech ends up dead in a bin nobody knows about. Tape has extreme density / cost ratio, but it suffers from the same problem Spinning drives are starting to run into, the time cost to save/retrieve a random bit from the media.
And this is why SSDs are going to relegate HDD to the dustbin of history ... eventually. The time cost for retrieving random data bits is too high. Long Term storage / backups is what tapes were known for, but even now, you're starting to see the same argum
Re: SSDs (Score:3)
I'm certified in NetBackup 7.5 for Unix and I'm a Dell EMC certified specialist for Data Domain 2.0
I manage 12,000 backup jobs per day, protecting 14 PB of data, with over 100 PB of protection capacity (to restore every copy of retained backup history).
I recently finished a 4 year long project to eliminate tape for all corporate system backups.
Purpose built backup appliances, object storage preceded by a dedupliaction engine, and even thin provisioned primary storage are all more effective and more efficien
Re: (Score:2)
Re: (Score:1)
Speaking only in the context of large data center storage arrays: SSDs have a finite number of writes, and cost 4x-10x what HDDs do per terabyte. OTOH, SSDs require less power, space, and cooling and are an order of magnitude faster. When sizing your storage, you need to take all that into account (and periodic price swings due to natural disasters and other supply-chain problems) to see which is more cost-effective. And don't forget that good caching can mitigate HDDs' relative slowness.
In reality though,
Re: (Score:2)
Despite the limited writes I think anyone doing anything IO intensive except maybe storing raw audio/video/sensor data is using SSDs for their active data now, or at least a SSD cache big enough to be the same in practice. HDDs is for stale data but not so stale you want to archive them to tape. You can tell by the use of SMR and various other technologies that actually make them even shittier than they used to be, in exchange for higher capacity per buck. I guess reliability a bit too but cost has been alm
Re: (Score:2)
HDDs are currently available at $15.33 per Terabyte (https://www.amazon.com/Hitachi-HUA723030ALA640-UltraStar-7K3000-7200RPM/dp/B005QTSDDQ)
SSDs are currently available at $93.74 per Terabyte (https://www.amazon.com/ADATA-SU655-960GB-Internal-ASU655SS-960GT-C/dp/B07K1NB8XN)
That means that SSDs are currently 6.11x more expensive than HDDs. And when you're using multiple devices in RAID, especially as BackBlaze are for back
Re: (Score:2)
Re: (Score:2)
I have no problem keeping up with watching video, installing games, etc. I even
Re:CALLED IT SINCE FOREVER, SEAGATE IS SHITE. (Score:5, Interesting)
Seagate are built down to a price. They are significantly cheaper than Hitachi drives, but also significantly less reliable.
Backblaze uses a lot of Seagate drives, because for them the cost of failures is low. They rip the drive out and shove a new one in, the data re-duplicates and the service is unaffected.
For people with higher failure costs, like lengthy restores from backup or performance destroying RAID rebuilds, Hitachi makes more sense. They are also popular with OEMs because warranty repairs are expensive.
Re: (Score:2)
"significantly cheaper than Hitachi drives, but also significantly less reliable." - 375 out of 400~ drives that failed here were Seagates. That's retarded. Saving even 50% doesn't make that a wise datacenter investment.
Your logic failed, reformat your brain. This is around a 4000-5000% failure rate increase for a slight price drop. There is no application except skeet shooting targets where that kind of failure rate "saves money" for anyone.
Retarded.
Don't you have to know what the return rate is, too? I mean, if you have a datacenter you have full-time staff. You have to have people there ready to fix stuff all the time, but things aren't broken all the time. So you have slack time for doing things like replacing dead drives, and it might not actually cost you anything. So you have to know if you're exchanging the dead drives for new ones before you can make stupid comments about "reformat your brain."
You didn't even introduce all the required variable
Re: (Score:2)
Yes, your analysis is retarded. 78% of the drives were S
Re: (Score:2)
Extraordinary no. And totaling drives and drive hours for the respective categories to determine percentages is certainly math.
Seagate 8TB drives are about as good as any HGST product with a comparably sized dataset. If you're s
Re: (Score:1)
What I got out of this is that with the exception of the 6TB size, Seagate is the most likely to fail and appears to average near 1 out of ever 50 vs less than 1 in 100 for other drives. What's most shocking is the 12TB failure rate, as those are pretty new.
Re: (Score:2)
Re: (Score:2)
I think you are trolling, pretending to be dumb. It wouldn't matter if the failure rate were infinitely higher: if no drives ever failed except Seagate, it would still make sense to use them when the cost of failure is preferable to the cost of better drives.
HGST FTW! (Score:1, Informative)
HGST still continues to be the overall winner.
Even WD hasn't seemed to be able to ruin them... yet.
Re: (Score:3)
A WDC employee I know told me several years ago that HGST operates independently of WDC in terms of its manufacturing and engineering groups. I believe that continues to be true.
HGST drives are difficult to source in quantity, which is probably why Backblaze deploys as many Seagate drives as they do. I buy several hundred drives a year as well, which is not a huge amount, but it's vastly easier at any point for my distributor to ship out Seagate than Western Digital.
really ? (Score:2)
Re:really ? (Score:5, Informative)
Not just different drives, totally different environment. This is all big 3.5" HDDs sitting quietly in storage pods with good cooling. What survives life in a laptop depends more on how rugged it is and how well the OEM implemented cooling than the "native" lifetime. Though with the last price drop I think HDDs in laptops is a thing of the past, little point now unless you're looking at 1TB+ drives.
Re: (Score:1)
Laptops at a big box store are three months of supply chain behind ones ordered direct.
Of course the parts are cheap! Of course they have old tech.
They know their customers don't know the difference.
Re:really ? (Score:5, Informative)
There are only three HDD manufacturers - Seagate, Western Digital, and Toshiba. HGST (Hitachi) merged with WD. But regulators made it a condition that they had to sell Hitachi's 3.5" production facilities and sell/license the designs and technology to Toshiba. Previously Toshiba only made 2.5" HDDs, so this condition guaranteed there would be three manufacturers of 3.5" HDDs after the merger instead of two.
All the other branded HDDs you see - Fujitsu, Samsung, LaCie, etc. - simply buy a hard drive from one of these big three, repackages it, and sells it to you as "theirs."
And the vast majority of new laptops with HDDs I've been opening up have Seagate drives. There used to be a lot more WD drives a few years ago, but WD has had a terrible head parking issue on their 5400 RPM drives (which includes most of their 2.5" drives including their SSHDs) for about 5 years now. It causes games and even the mouse pointer to intermittently freeze for about a quarter to half second [tomshardware.com]. I think laptop manufacturers have finally figured out that the WD drive is the cause of all the complaints they've been getting about freezing and stuttering on their laptops, and they've been ditching WD drives for Seagate.
Statistically, to differentiate between a 4% annual failure rate and an 8% annual failure rate with a 95% level of confidence, you need a sample size of about 100 of both drive models. So really, it's only companies which use a large number of drives (like Backblaze) which can reliably produce these kinds of statistics. A repair shop can't because you're only seeing the failures. You'll see more of the popular drives coming in for repair simply because they're popular, not necessarily because they fail more. To calculate a failure rate, you need to know the number of failures AND the number which didn't fail.
Finally a good /. post that's not political (Score:2)
Thank you for finally posting something that lacks a political narrative. This is actually useful information.
Confidence intervals? (Score:5, Insightful)
As always, this is very interesting data, as it represents the fairly large population that is needed for better statistical analysis. However, Backblaze then fails to present the data along with the measure of statistical confidence that is the entire reason for trusting its numbers more than smaller sample sizes.
With the confidence intervals, the conclusions change at least for some models. As expected, the drive models with small sample sizes have huge confidence intervals. For example, the HGST HUH721212ALE600 has an estimated AFR of 2.60%, but the 95% confidence interval ranges from 0.07% to 14.48%, so there simply isn't enough collected data to be significant. In contrast, the Seagate ST12000NM0007 also has a high AFR of 2.22% but with a much tighter confidence interval of 1.91% to 2.22%. So, while the AFR for these two drives appear to be high, only the Seagate drive can be shown to have a high AFR with confidence.
Unfortunately, only 5 out of 15 drive models have upper confidence interval ranges that are less than 100%. So, the conclusions from this data need to be viewed in that context.
Lifetime Hard Drive Stats (Score:2)
Data from Backblaze are very relevant, they show actual figures for HDD reliability. Interestingly, this is their first plot showing the (average) age of the disks (in months, I believe) together with the Annualized Failure Rate (AFR).
I've bothered to plot the numbers of that chart, and interestingly they do not resemble a clear bathtub curve [wikipedia.org]. However, the data in the Age column is the average for all disks from the same model; probably it is mixing both old and new disks, except for very low values. I woul