Reliability issues on 6TB internal HDDs?

[goretsky Supervisor]

Hello,

 

BackBlaze just released their 2018 HDD Stats report earlier today:  https://www.backblaze.com/blog/hard-drive-stats-for-2018/

 

Perhaps you will find this of interest.

 

Regards,

 

Aryeh Goretsky

[DevTech]

16 minutes ago, goretsky said:

Hello,

 

BackBlaze just released their 2018 HDD Stats report earlier today:  https://www.backblaze.com/blog/hard-drive-stats-for-2018/

 

Perhaps you will find this of interest.

 

Regards,

 

Aryeh Goretsky

"As of December 31, 2018, we had 106,919 spinning hard drives. Of that number, there were 1,965 boot drives and 104,954 data drives. This review looks at the hard drive failure rates for the data drive models in operation in our data centers. In addition, we’ll take a look at the new hard drive models we’ve added in 2018 including our 12 TB HGST and 14 TB Toshiba drives. Along the way we’ll share observations and insights on the data presented and we look forward to you doing the same in the comments."

 

106,919 hard drives is a lot, but nothing compared to the big cloud players, but THIS IS ALL WE HAVE!

 

With all the hard drives Amazon is using in AWS and Microsoft in Azure and Google in GCP and well Google everywhere, Backblaze is the ONLY large user of disk drives decent and good enough to share reliability data with humanity!

 

 

[DevTech]

"Finally in Q4, we added 1,200 Toshiba 14 TB drives (model: MG07ACA14TA). These are helium-filled PMR (perpendicular magnetic recording) drives."

 

I would have a difficult time trusting a Helium filled drive for longevity since Helium is so small it will leak through any metal over time...

 

"Speaking of HGST drives, in Q4 we added 1,200 HGST 12 TB drives (model: HUH721212ALN604). We had previously tested these drives in Q3 with no failures, so we have filled a Backblaze Vault with 1,200 drives. After about one month we’ve only had one failure, so they are off to a good start."

 

As a group, the Hitachi drives have the lowest failure rate over time. So why doesn't Backblaze buy Hitachi drives over Seagate which have had some of the highest failure rates? The reason is that a drive failure is a monetary event to them. They have multiple copies of any data so all a drive failure means is that it needs to be replaced and another one purchased. So if Seagate drives have a higher failure rate but are a lot cheaper to buy, then they are more attractive overall. 

 

For an individual, the trade-off is different and in most cases trading extra money for more reliability is a very good deal. The problem has been that there has been no way to be certain about reliability...

 

On a quick look, the most reliable drives right now 6 TB or larger appear to be:

 

1. Hitachi HUH721212ALN604 12 TB - 0.51 %

 

2. Seagate ST10000NM0086 10 TB - 0.33 %

 

Note that if you make a purchase based on that data, the drive model number must match exactly for you to benefit from the reliability info.

 

As a consumer and not an objective mathematician, failure rates of 1/2 of 1% are bad enough that 1% is just uncomfortable to think about - 1 in a 100 in any given year is a horrible lotto for precious data...

 

Which leads me to completely rethink the WD Red 6 TB with a horrible failure of 5.5% to 2% averaged at 4% - WTF.

 

So, I still feel the more expensive construction of the WD Black should actually result in good reliability (and fantastic performance is built in) but for new "Mass Storage" drives, the Hitachi 12 TB or Seagate 10 TB are the ones to get.

 

If the Hitachi 4 TB is still available it is 0.36% and suitable.

 

Hmm maybe that will be hard to source - $70 refurbished at Newegg:

 

Hitachi HMS5C4040BLE640 $70

 

https://www.newegg.com/Product/Product.aspx?Item=9SIA5AD8N69185&Description=HMS5C4040BLE640&cm_re=HMS5C4040BLE640-_-9SIA5AD8N69185-_-Product

 

Seagate ST10000NM0086 $400

 

https://www.newegg.com/Product/Product.aspx?Item=9SIA1K68GH1882&Description=ST10000NM0086&cm_re=ST10000NM0086-_-9SIA1K68GH1882-_-Product

 

Hitachi HUH721212ALN604 448 in pounds, (but in 37 days it might be 2,000,000 inflated May Bucks...)

 

https://www.amazon.co.uk/HGST-Ultrastar-HE12-12TB-HUH7-21212ALE604-24x7-8-9-3-5-Inch/dp/B071W4BVVT/ref=pd_lpo_sbs_107_t_2?_encoding=UTF8&psc=1&refRID=P4HY7M2K4H5H9N1V1H48

 

For reference:

 

WD Black 6 TB $230

 

https://www.newegg.com/Product/Product.aspx?Item=234-000G-000W5&Description=wd black&cm_re=wd_black-_-234-000G-000W5-_-Product

 

 

 

[Mindovermaster Moderator]

15 minutes ago, DevTech said:

<snip>

very interesting....

[Technique]

Yeah I gather that all hard drives will die. No doubt every single component in a PC will 1 day fail as well. 

 

Anyway I read the stuff on my last post on user reviews on Amazon. I didn’t insta-swallow it, hence why I came here to check. 

 

Thanks. 

[DevTech]

16 hours ago, Technique said:

Yeah I gather that all hard drives will die. No doubt every single component in a PC will 1 day fail as well. 

 

Anyway I read the stuff on my last post on user reviews on Amazon. I didn’t insta-swallow it, hence why I came here to check. 

 

Thanks. 

It is darned hard to select a drive from the limited types that Backblaze uses, but it is the ONLY source of information on reliability and you can see from their data that there are some bad drive models.

 

If pressed, I would even have to say my preference for WD Black based on tech analysis does not hold water at all compared to actual reliability data.

 

If you can't get a "good" drive from the Backblaze list, then get a WD Black and if you can't get that, go for the Seagate IronWolf since WD Red based on the Backblaze data is just too unreliable, but maybe the IronWolf isn't too bad based on some of their newer Seagate data, but we will never know for sure...

 

 

 

[DevTech]

39 minutes ago, Technique said:

Yeah I gather that all hard drives will die. No doubt every single component in a PC will 1 day fail as well.

I missed that. It is better to think of every single component with separate reliability profiles.

 

I'd say that reliability goes in this order:

 

- keyboards and mice fail a lot

- hard drives fail more often than any other "key" component

- power supplies don't do well if they are not top-of-the-line models

- mobos are next, typically the high current power to the CPU chip.

- monitors

- I don't think people treat RAM very well because I hear of a lot of failed RAM, but I don't experience it ever. Always goes completely obsolete before failing

- CPUs NEVER fail unless mistreated.

 

So, hard drives are a real concern, the rest of it not so much...

 

[Daniel F. Administrators]

5 hours ago, DevTech said:

So, hard drives are a real concern, the rest of it not so much...

 

I think anything that is mechanical should top the list and anything used/run repeatedly should top the list (kb & mice fans) (HDD depends on how many hours its quoted. 25,000 hours (3 years) normally being the minimum. Usually quoted in their warranty).

Anything memory like ssd's have a finite read/write cycle and will then become unusable.

Anything overclocked or subject to excessive voltage/heat - for obvious reasons 

[DevTech]

30 minutes ago, Daniel F. said:

 

I think anything that is mechanical should top the list and anything used/run repeatedly should top the list (kb & mice fans) (HDD depends on how many hours its quoted. 25,000 hours (3 years) normally being the minimum. Usually quoted in their warranty).

Anything memory like ssd's have a finite read/write cycle and will then become unusable.

Anything overclocked or subject to excessive voltage/heat - for obvious reasons 

I had left SSDs off my little ad-hoc list due to complexity.

 

SSD's share one characteristic with keyboards and mice in that they wear down the more you use them. Current generation SSDs will almost certainly make it past the warranty period but then it gets tricky.

 

Once spare cells run down to ZERO, the drive is dead. Personally, I would wish they had been smart enough to just mark the sector involved as an unrecoverable error and then let the O/S remap sectors like it would do for a hard drive. You could then rescue the drive by re-partitioning it, even cut to half-size since spare cells are zero at that point and then the huge new empty partition becomes new spare cell repository for the firmware. It's not hard to imagine the industry doing a standard API between SSD firmware and OS to make that happen...

 

It's Neowin tech people here and some percentage of techies like to eke out every bit of value and lifetime from equipment and in the past most devices (if initially purchased from quality manufacturers) could be expected to survive well past the warranty period by a few multiples even. Hit and miss of course, but still the majority of quality stuff really lasts well...

 

With SSDs, depending on how they are used, there is essentially a countdown clock at end of warranty. Every single write to the drive brings it closer to Drive Cemetery.

 

SSDs come from the factory with a spare cell allotment of 5% to 15% of the drive depending on the FLASH tech involved. I always add another 10% to that on every drive in the hope my SSDs will last like some of my hard drives have...

 

I have already murderd one drive by writing to it too much (first gen doing giant C++ compiles, but still a shock) which opened my eyes to  this new class of all-electronic device that somehow had the feel of a mechanical device getting wear and tear!