Hard Drive Reliability Stats for Q1 2015

[ez24]

Hi

I came across this link (from another link on the post on backups)

https://www.backblaze.com/blog/best-hard-drive/

I think I would be afraid if I had a Seagate. 

I wish I had seen this 2 days ago because I just bought 4 - 4tb WD drives ( 9% failure) because they were $15 each cheaper than Toshiba ( 0 % )   

Here is a link to their raw data if someone likes this sort of thing

https://www.backblaze.com/hard-drive-test-data.html


Maybe someone can explain this to me:

                The Toshiba 5 TB  has a   "95% Confidence Interval"  of 106.9 %    ?

[Balaur]

Maybe someone can explain this to me:

                The Toshiba 5 TB  has a   "95% Confidence Interval"  of 106.9 %    ?

No events (failures) and very short average age (0.1 years). 
(The chi is strong with this one )

Simply put, a failure may happen just after the observation period, putting the risk at a very high level.
On the other hand, the device may function indefinitely without failures =>  very low risk.
There is no indication yet on what would happen.

If you already had thousands of events, then, statistically speaking, the devices are much more probable to behave like the average, leading to a very small confidence interval centered on a very well justified average.

[German_EE]

Now you know why I avoid Seagate.

[nuno]

As far as I understood it's the 1st quarter they're using the Toshibas marked 0.0%... (average age, 0.1years) [as Balaur already mentioned]
The only other they have with 0.1 years is the Seagate 6 TB SATA 3.5, which has 2.41% failure rate, but, they have 495 of these Seagates while only 90 of all the 0.1 year Toshibas combined.

[dr.diesel]

Now you know why I avoid Seagate.

In a 15 * 4TB array, I've lost 5 drives in the past 2 years.  (All Seagate ST4000DM000)   

[MagicSmoker]

I had to chuckle when I saw that the once-infamous Hitachi/IBM "DeathStars" (aka DeskStars) have been one of the more more reliable brands in their experience. Sort of like how Hyundai went from trash (in the late 80's) to flash (now).

[Fsck]

I had to chuckle when I saw that the once-infamous Hitachi/IBM "DeathStars" (aka DeskStars) have been one of the more more reliable brands in their experience. Sort of like how Hyundai went from trash (in the late 80's) to flash (now).

that was only one generation of epic fail.
I'm actually kind of disappointed at the WD Reds.

[coppice]

I like Hitachi. They are a bit more expensive, but you get what you pay for. It's been that way for years. They are good about warranty replacements too.

The very different failure rates of the recent HGST and WD drives in that table suggest that WD are still operating the HGST business separately. I doubt that will last. HGST's failure rates went through big peaks and troughs when they were IBM. I haven't used their products much under the Hitachi name.

[oxcart]

From our numbers(though smaller than BlackBlaze), the landscape is the same.
Seagate quality is also variable: so you might get an ST in a bashed laptop that still works fine. But you'd have to be lucky.

[sync]

Now you know why I avoid Seagate.

In a 15 * 4TB array, I've lost 5 drives in the past 2 years.  (All Seagate ST4000DM000)   

Sorry, but they are the wrong type of disks for this. Only rated for 2400h/year. Get some enterprise disk for it.

[dr.diesel]

Sorry, but they are the wrong type of disks for this. Only rated for 2400h/year. Get some enterprise disk for it.

My stacks of ESs are no better.

[sync]

I had to chuckle when I saw that the once-infamous Hitachi/IBM "DeathStars" (aka DeskStars) have been one of the more more reliable brands in their experience. Sort of like how Hyundai went from trash (in the late 80's) to flash (now).

that was only one generation of epic fail.

It was two generations at least. First the DTLA series and it's successor (IC35L.. something). With the IC35L.. series we had about 40% failure rate in the first months. We replaced them all.

[sync]

Sorry, but they are the wrong type of disks for this. Only rated for 2400h/year. Get some enterprise disk for it.

My stacks of ESs are no better.

Interesting. Which models? I have good experience with the Constellation ES series. Maybe 100 drives total.

[dr.diesel]

Variety over the years, and it's not a comprehensive study either, as many were covered by the SAN contracts.

I quit using the ESs at home years ago, rely on RAID6 and onsite spares, just swap them out, RMA, move on.

[BloodyCactus]

(knows its going to be backblaze before opening the thread... yep!)

are backblaze still using cheap consumer level drives instead of NAS/Enterprise level drives for their mass storage? 

[ez24]

are backblaze still using cheap consumer level drives instead of NAS/Enterprise level drives for their mass
storage?

Yes they say this and the stats are for consumer drives like I  use (I am a consumer).  It would be interesting to compare the stats with the "other" drives.

[oxcart]

(knows its going to be backblaze before opening the thread... yep!)

are backblaze still using cheap consumer level drives instead of NAS/Enterprise level drives for their mass storage? 

Google made a good point about this . Being the largest(?) hdd buyer worldwide, they simply found out that the best solution is to whack in the cheapest drive  on the market, providing of course that you do not rely on the integrity of a single drive.

Basically, a big embarassement for NAS/Enterprise drive makers. 

Of course, on my private, few-bays NAS i absolutely won't  slap in those cheap ass drives, as my data integrity would be much, much, much more at risk. 

[BloodyCactus]

(knows its going to be backblaze before opening the thread... yep!)

are backblaze still using cheap consumer level drives instead of NAS/Enterprise level drives for their mass storage? 

Google made a good point about this . Being the largest(?) hdd buyer worldwide, they simply found out that the best solution is to whack in the cheapest drive  on the market, providing of course that you do not rely on the integrity of a single drive.

Basically, a big embarassement for NAS/Enterprise drive makers. 

Of course, on my private, few-bays NAS i absolutely won't  slap in those cheap ass drives, as my data integrity would be much, much, much more at risk. 

yep. you can use the cheapest drive around, but you get what you pay for!

if you can afford to use a drive for a week and replace it, great.

my Segate NAS branded drives have run 24x7 without a hitch (I dont have the spare money nor do I want to spend the time rebuilding my arrays!) or needed replacement.

[GoneTomorrow]

my Segate NAS branded drives have run 24x7 without a hitch (I dont have the spare money nor do I want to spend the time rebuilding my arrays!) or needed replacement.

Yeah same, I've had 4x 3TB Seagate NAS in an array for the last 2 ish years. They're all showing about 18,000 power on hours, and nothing untoward in the smart data. Array not reporting any errors either.

HDD's always amaze me; even with only 18,000 hours the platters have spun around a stupid number of times: roughly 6,372,000,000 (18000 hours * 60 minutes * 5900? rpm) Dem bearings  O0

[urbis]

I've never had a Seagate drive fail, I've got 6 consumer drives running in a NAS enclosure that have been there a good while.

Everything's backed up and RAID though!

[edavid]

(knows its going to be backblaze before opening the thread... yep!)

are backblaze still using cheap consumer level drives instead of NAS/Enterprise level drives for their mass storage? 

They did try enterprise drives, but the reliability was worse 

https://www.backblaze.com/blog/enterprise-drive-reliability/

[kc8apf]

The reliability varies quite a bit from model to model regardless of vendor.  While Google long ago used the cheapest consumer HDD, they haven't for over 6 years.  They buy server-class (also called nearline) drives from multiple vendors.  Every other year, some drive model will have a design problem that leads to early failure.  At that scale, it's more important to spread out across multiple vendors and models to mitigate that model-specific failure so it doesn't become an epidemic.

HGST tends to build their consumer drives by cost reducing their server-class designs.  WD and Seagate tend to design consumer and server drives independently but use them both for introductions of new technologies.  Toshiba hasn't had server-class drives until just a few years ago when they attempted to scale up their consumer drives into something more reliable.

There are lots of subtle effects that can affect what gets reported as failure rates.  Many, many times the failure is a read failure and that sector is lost but the a write to that sector will force a reallocation and the drive will be perfectly fine.  Contrary to common belief, a few (~15) bad sectors on a drive is not an indicator that the drive will truly die soon.  With the bit patterns being so dense nowadays, sector failures are inevitable.  That's why each drive is tested by the manufacturer and given a permanent table of known bad sectors.

Bottom line: if you want rock-solid drives, HGST is expensive but well built.  Seagate is nearly as good and has some other useful features (security features, etc) at a slightly better price.