Backblaze Hard Drive Stats for 2017

R

rgMekanic

[H]ard|News
Joined
May 13, 2013
Messages
6,943
Backup company Backblaze has released it's hard drive data from 2017. During Q3 and Q4 Backblaze removed nearly all the 3TB drives it had in service, and began installing 10TB and 12TB drives in their place. One thing that has been true year after year, HGST drives are pretty reliable.

I look forward to this report every time. Because of the infrastructure Backblaze has built up it gives a great look into the reliability of these hard drives in a fairly harsh environment. Currently Backblaze has 91,305 data drives in operation, crazy.

At the end of 2017 we had 93,240 spinning hard drives. Of that number, there were 1,935 boot drives and 91,305 data drives. This post looks at the hard drive statistics of the data drives we monitor. We'll review the stats for Q4 2017, all of 2017, and the lifetime statistics for all of the drives Backblaze has used in our cloud storage data centers since we started keeping track. 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.
 
WhoMe said:
Yes, they did, they sold a lot more of their drives to Backblaze than any other brand.
Click to expand...

..and they <consideration> sold them their consumer drives. Enterprise and high application expressively manufactured for high vibration, heat and 24x7 environment drives are too expensive.

Consumer drives in such an environment makes it seem that they are unreliable. As much as I appreciate the backblaze numbers, gotta consider that better drives would perform better.

- I am not a seagate investor :)
 
WhoMe said:
Yes, they did
Click to expand...
But did they?

They sell a lot less drives to everyone else though! Because everyone else doesn't have other people paying for the failed drives so they have to buy the more reliable options.
 
Khahhblaab said:
Consumer drives in such an environment makes it seem that they are unreliable. As much as I appreciate the backblaze numbers, gotta consider that better drives would perform better.
Click to expand...
I believe they already addressed this previously. They have tested the enterprise drives against none enterprise drive and they were either worse or within the margin of error that it was far better CBA to buy the cheaper drives to have the near exact same performance / failures for cheaper.
 
HoffY said:
I believe they already addressed this previously. They have tested the enterprise drives against none enterprise drive and they were either worse or within the margin of error that it was far better CBA to buy the cheaper drives to have the near exact same performance / failures for cheaper.
Click to expand...

Possibly.

But a drive that is expected to be in a high vibration, high heat environment will be engineered to be there. A consumer drive isnt expected to have to perform in the same environment. Maybe as simple as adding a gasket of rubber between the metal to metal contact points.

As much as I dont want to remember, I remember when I had to buy a scsi drive for a failing one in a electron microscope......overnight.

Supplier cost, 9 thousand!

Once the drive was in my hands, it was a simple scsi drive. Lets see if this will work. Maintenance guy drove to walmart and bought a seagate drive.

From backup, copied data and we put it in the microscope.

Yep! It worked. $350.

So there are comparable's to a basic drive and one 'manufactured' to be in high stress environments. But 9K?

I still, even with this example feel that the expectation of a drive being in an environment that is 'not' consumer will perform better.

If backblaze is going to have to pay over 8K more for a drive specifically for a server environment I can see how the ordinary drives is a better option. But the hardened drive will have reasons for its usage.

A tank cant be compared to an SUV.................but we do.
 
Khahhblaab said:
But a drive that is expected to be in a high vibration, high heat environment will be engineered to be there.
Click to expand...
But this is what you are inccorectly making assumptions on. The HDDS are not made to some ridiculous "survive in space" conditions. They are virtually all the same hardware with a few tweaks to firmwares and then sold at 5x the price with mostly marketing buzz words thrown in to fool the idiots. When the drives get tested side by side in the real world.. they offer no such claimed advantages. The differences are so small that being "engineered for high vibration blahb blah blah" is just B.S. Henc.e.. failed CBA and they use consudmer drives.

Blackblaze aren't the first to come to this conclusion either. I have been highly suspicious of this type of thing for decades and it always ends up being proven with unbiased user tests all whilst people yell "you're a conspiracy theorist" or "racist" to try and dismiss what the actual proof shows outside of the marketing spinners.

Just look way back at the soft quadro days for yet another example of such "it costs so much in R&D and its 'engineered' for pro's thats why we charge 4x as much, promise!" and yet 5 minutes of software changes and it runs just as fast as the "fully engineered for tougher terrain" models!

Its just B.S. and people need to wake up and wise up about getting ripped off so much. This is why we have so much Boiling Frog Syndrome happening nowadays because gullible masses.
 
I will say this. HGST REALLY hit the jackpot with their 4TB drives!

Those are some fucking CRAZY-LOW failure rates...
 
Last edited by a moderator:
Khahhblaab said:
A tank cant be compared to an SUV.................but we do.
Click to expand...
Of course you can compare a tank to an SUV. But you miss the point.. its not what you compare.. its the reason for comparing them and then the results one may find, that matters.

If your criteria is to blow things up but you have a budget of 5000$, comparing those two may be a wise choice because you can go buy a rocket launcher and a cheap SUV and bingo... comparison resulted in the ability to make better decisions because one wasn't so narrow minded. ;)
 
Chas said:
I will say this. HGST REALLY hit the jackpot with their 4GB drives!

Those are some fucking CRAZY-LOW failure rates...
Click to expand...
I hope they do just as good a job on the 8TB drives as i just bought two to go with 6 other WD 8TB drives. Going to be raining all weekend here so perhaps i will pull down the puter and install them. :)
 
HoffY said:
I hope they do just as good a job on the 8TB drives as i just bought two to go with 6 other WD 8TB drives. Going to be raining all weekend here so perhaps i will pull down the puter and install them. :)
Click to expand...


Unfortunately, they don't have enough of the 8TB units to really make a good call. Granted, they had a couple failures early on. But none in the last 2 years. So I'm willing to chalk those dead drives up to "bathtub curve". It seems to indicate that the HGST drives are, overall, a quite durable platform.
 
Khahhblaab said:
..and they <consideration> sold them their consumer drives. Enterprise and high application expressively manufactured for high vibration, heat and 24x7 environment drives are too expensive.

Consumer drives in such an environment makes it seem that they are unreliable. As much as I appreciate the backblaze numbers, gotta consider that better drives would perform better.

- I am not a seagate investor :)
Click to expand...

That may have been true for their 4TB drives, but their 8, 10 and 12TB seagate drives are all enterprise.

st12000nm0007
st10000nm0086
st8000nm0055
 
I have been giving Toshiba a go lately as they seem similar to HGST and had decent success with 1 failure in 2yrs drive power on time. All my customers use HGST with no failures as per usual.

Meanwhile my 4 Seagate 4TBs are excellent doorstops after failing with no warning.
 
Khahhblaab said:
..and they <consideration> sold them their consumer drives. Enterprise and high application expressively manufactured for high vibration, heat and 24x7 environment drives are too expensive.

Consumer drives in such an environment makes it seem that they are unreliable. As much as I appreciate the backblaze numbers, gotta consider that better drives would perform better.

- I am not a seagate investor :)
Click to expand...


IIRC, backblaze has compared enterprise "grade" drives and consumer "grade" drives and found basically little to no difference in drive reliability in their environment.
 
Khahhblaab said:
Possibly.

But a drive that is expected to be in a high vibration, high heat environment will be engineered to be there. A consumer drive isnt expected to have to perform in the same environment. Maybe as simple as adding a gasket of rubber between the metal to metal contact points.
Click to expand...

Unless there is a real actual physical difference between the drives in the different lines, you aren't gaining anything with the enterprise drive. And the stark reality is they make 1 drive type and then use different firmware. It is just more cost effective to have all the hardware pieces the same as much as possible. It also allows you to shift the drives into a wider variety of markets. WD makes one 8TB drive model and they can sell it into the AV market, security market, consumer market, prosumer market, prosumer raid market, enterprise market, usb market, usb raid market, etc.

The factory doesn't have to make constant changes to the line to support slightly different variations of the same base design, all that changes is the firmware and the sticker.
 
Khahhblaab said:
Possibly.

But a drive that is expected to be in a high vibration, high heat environment will be engineered to be there. A consumer drive isnt expected to have to perform in the same environment. Maybe as simple as adding a gasket of rubber between the metal to metal contact points.

As much as I dont want to remember, I remember when I had to buy a scsi drive for a failing one in a electron microscope......overnight.

Supplier cost, 9 thousand!

Once the drive was in my hands, it was a simple scsi drive. Lets see if this will work. Maintenance guy drove to walmart and bought a seagate drive.

From backup, copied data and we put it in the microscope.

Yep! It worked. $350.

So there are comparable's to a basic drive and one 'manufactured' to be in high stress environments. But 9K?

I still, even with this example feel that the expectation of a drive being in an environment that is 'not' consumer will perform better.

If backblaze is going to have to pay over 8K more for a drive specifically for a server environment I can see how the ordinary drives is a better option. But the hardened drive will have reasons for its usage.

A tank cant be compared to an SUV.................but we do.
Click to expand...

This is like going to the dealership to buy a fan belt for a car, you can get the same or better belt at Autozone for 1/3 the price or less, even made by the same manufacturer with the same part number. We have an industrial microwave at the lab for processing samples, it has a rectifier that goes out on a regular basis. The manufacturer of the microwave sells that part for $200, but we can source the exact same part(peeled the manufacturer label back and read the OEM part number underneath) from a local electronic supplier for $15. No difference in the part or it's performance or lifetime. Same thing happens with all of our analytical instruments such as mass spectrometers and such. Anything that is a normal off the shelf part such as fuses and switches cost 10x if bought from the manufacturer.

I wish I could get our IT guy at work to understand these things. We have many Dell desktops that run the equipment and have drive failures all the time since the lab is full of acid and solvent fumes, but the older computers just chug along. He will not replace a drive with anything other than what Dell will sell him, even if they continue to fail. Of course he is a database guy, not a hardware geek like me, but he is the type that always takes his car back to the dealership for repairs too.
 
aaronspink said:
Unless there is a real actual physical difference between the drives in the different lines, you aren't gaining anything with the enterprise drive. And the stark reality is they make 1 drive type and then use different firmware. It is just more cost effective to have all the hardware pieces the same as much as possible. It also allows you to shift the drives into a wider variety of markets. WD makes one 8TB drive model and they can sell it into the AV market, security market, consumer market, prosumer market, prosumer raid market, enterprise market, usb market, usb raid market, etc.

The factory doesn't have to make constant changes to the line to support slightly different variations of the same base design, all that changes is the firmware and the sticker.
Click to expand...
^This
Toms hardware even spells it out and says some of the "pro" drives fail more frequently, but at least you pay more for the privilege.
http://www.tomshardware.com/news/backblaze-q417-hdd-reliability-report,36452.html "Many enterprise drives are certainly not showing standout durability, while many consumer drives are showing surprising reliability. "
 
Chas said:
I will say this. HGST REALLY hit the jackpot with their 4TB drives!

Those are some fucking CRAZY-LOW failure rates...
Click to expand...

Long gone are the IBM days of rightfully being called 'deathstars'.
 
aokman said:
I have been giving Toshiba a go lately as they seem similar to HGST and had decent success with 1 failure in 2yrs drive power on time. All my customers use HGST with no failures as per usual.

Meanwhile my 4 Seagate 4TBs are excellent doorstops after failing with no warning.
Click to expand...

The only drives that have ever failed in my personal use have been Seagate. Well, except one WD which I killed by accidentally plugging a cheap chinese Molex power connector backwards. It shouldn't have been possible to go in upside down but it did :)
 
B00nie said:
Long gone are the IBM days of rightfully being called 'deathstars'.
Click to expand...
I had 2 of those in raid 0 running either xp or 2000, dang it was fast while it lasted. Ran a chkdsk just for maintenance and it turned into a blue screen full of numbers and symbols marching by as the scan went on (never seen before or since), shortly there after the raid failed with the death of a drive and then not too much later the other drive checked out. happy happy joy joy
 
I stopped buying Seagate drives like 5 years ago when I threw 10 of them in the trash that came out of my Primary and Backup media servers. They had tons of surface errors and media defects. Many were under warranty but what good is that when Seagate sends me someone else's defective used drive time after time. Solution? I only buy HGST spinners and Samsung SSDs. I have never had a HGST drive fail. I have like 30 of them that are 4, 5, 6, 8 and now 10TB in size. I use HDSentinel to monitor all my drives and report any new defects or other issues. I have only had one Samsung SSD fail in like over a hundred I have purchased. It was an 840EVO. They overnight-ed me a new one, not a refurb.
 
So a few years ago (2011? Can't remember) I got a great deal on a bulk buy with some friends on 1.5TB seagates.

Most of them died within a few years. I thought I had replaced all mine, but i recently went to check:

Code: 
# smartctl -i /dev/ada2
smartctl 6.6 2017-11-05 r4594 [FreeBSD 11.1-RELEASE-p4 amd64] (local build)
Copyright (C) 2002-17, Bruce Allen, Christian Franke, www.smartmontools.org

=== START OF INFORMATION SECTION ===
Model Family:     Seagate Barracuda 7200.11
Device Model:     ST31500341AS

# smartctl -A /dev/ada2 | grep "Power_On"
  9 Power_On_Hours          0x0032   051   051   000    Old_age   Always       -       43572

1815 days!

Code: 
# smartctl -i /dev/da2
smartctl 6.6 2017-11-05 r4594 [FreeBSD 11.1-RELEASE-p4 amd64] (local build)
Copyright (C) 2002-17, Bruce Allen, Christian Franke, www.smartmontools.org

=== START OF INFORMATION SECTION ===
Model Family:     Seagate Barracuda 7200.11
Device Model:     ST31500341AS

# smartctl -A /dev/da2 | grep Power_On
  9 Power_On_Hours          0x0032   034   034   000    Old_age   Always       -       57918

2413 days!

Code: 
# smartctl -i /dev/da6
smartctl 6.6 2017-11-05 r4594 [FreeBSD 11.1-RELEASE-p4 amd64] (local build)
Copyright (C) 2002-17, Bruce Allen, Christian Franke, www.smartmontools.org

=== START OF INFORMATION SECTION ===
Model Family:     Seagate Barracuda 7200.11
Device Model:     ST31500341AS
# smartctl -A /dev/da6 | grep Power_On
  9 Power_On_Hours          0x0032   055   055   000    Old_age   Always       -       39951
1664 days.

Not bad for Seagate. I have another 2, 7200.12 that are > 32k hours.

I have some still working 74G raptors that broke 70k power on hours and I retired them because they were simply too small to be usable (esp when 120G ssds are $50 now...)
 
B00nie said:
Long gone are the IBM days of rightfully being called 'deathstars'.
Click to expand...

I'm still sympathetic to IBM on that front.
They DID spec the "Deathstar" series drives for, IIRC, 8 hour duty cycles and never intended the drives to be up 24x7x365.
 
How is Seagate still in business?

I got a new Tivo Bolt. It uses 2.5" hard drives - how stupid. The popular up grade drive in the Tivo forums was a 6TB Seagate. I was apprehensive, but went ahead and pulled the trigger.

It's horrible. It's the first time in 10 years I have had a Tivo stutter or glitch because the hard drive couldn't keep up. Seagate quality once again shining through. Never again!

I'm about to get a 6TB HGST 3.5" and an external power supply for it and just lay it next to the bolt with a different SATA cable. The Tivo is in a cabinet so the hard drive "out in the open" isn't a big deal. It's what I should have done in the first place.
 
On the topic of enterprise drives: what you pay for is the warranty. The physical hardware is the same.
 
FWIW....

Back in July, 2016, I purchased EIGHT HGST 5TB NAS drives for $150 each for Vaultron, my FreeNAS server, and configured in a RAIDZ2 configuration. One of the drives suffered infant mortality, and died within 48 hours. It was RMA-ed, and replaced without question with a new drive. After replacing the drive and resilvering the array, my FreeNAS server has been chugging along just fine. I've used 40% of the available capacity, with most of my storage being used by the games archive (23% of used capacity) and backups (28% of used capacity). I do backup my FreeNAS server to several external drives, with my games archive going to a 8TB Seagate external (Thanksgiving sale) and the rest going to either 2 TB or 4 TB Western Digital externals.

It should be noted that my FreeNAS server and my six-year-old laptop are the only computers on my personal network that still have physical drives with the rest having SSD drives. And, that's not entirely true for the laptop.... since the laptop has two 2.5" drive bays, the primary drive is a 500GB SSD while the secondary drive is a 1TB HD.

What really sucks is that we have two primary manufacturers of hard drives: Seagate and Western Digital. (HGST is part of WD.) There is no good reason for anyone to go into the hard drive manufacturing business.
 
seems like they're buying a lot more HGST drives than they use to. I'm surprised they aren't buying out WD drives from Best Buy. That said, seems like most are pretty solid, except the M001 and M005 Seagate models, which are scary bad
 
nilepez said:
seems like they're buying a lot more HGST drives than they use to. I'm surprised they aren't buying out WD drives from Best Buy. That said, seems like most are pretty solid, except the M001 and M005 Seagate models, which are scary bad
Click to expand...

The only time they really went retail was when general availability of drives was scarce. It make significantly more sense for them to buy directly in bulk at their volumes. When you are putting 450 drives per rack and multiple racks per week into service, you are at a volume level higher than any retail location.
 
aaronspink said:
The only time they really went retail was when general availability of drives was scarce. It make significantly more sense for them to buy directly in bulk at their volumes. When you are putting 450 drives per rack and multiple racks per week into service, you are at a volume level higher than any retail location.
Click to expand...
I'll take your word for it. I thought at one point they were shucking external drives, because they were cheaper.
 
nilepez said:
I'll take your word for it. I thought at one point they were shucking external drives, because they were cheaper.
Click to expand...

no, they were shucking externals because the entire market was in a capacity shortage with prices going sky high. In late 2011 there was massive flooding in Thailand that took a major WD plant off line along with many component suppliers which ended up affecting Seagate as well. WD lost >60% of its capacity because of the flood. Seagate's factories were spared flooding but their suppliers got hit. Overall, supply was basically reduced by 30-40% fo 12+ months. Prices increased 100% rather quickly and stayed there for ~15-18 months as even once everything was back online 9-12 months later, there was massive pent up demand. This was very early inthe days of backblaze and they had to scramble to do whatever they could to get HDs so started to buy externals wherever they could find them since they were a bit cheaper. For a while they basically had a bounty program for all friends and family. Once the crisis was over, supplies returned to normal, and prices became sane again, they went to buying in bulk directly. But back during the crisis period, you basically could even put in an order. All the big OEMs had first dibs, were paying 2x, and still could get enough drives to meet their internal demand.

To give an idea of the impact, WD 2TB green drives were going for ~70 before the flood, shot up and peaked at 250 and eventually settled at around 170 for the duration. Seagate Barracuda 1TBs went from 120 to ~200. And that was IF you could find stock.
 
HoffY said:
But this is what you are inccorectly making assumptions on. The HDDS are not made to some ridiculous "survive in space" conditions. They are virtually all the same hardware with a few tweaks to firmwares and then sold at 5x the price with mostly marketing buzz words thrown in to fool the idiots.
Click to expand...

Often the primary differentiator between retail drives and enterprise drives certified for application X is that someone paid to do lab certification of the performance of the drive, which costs real money up front, so they claw it back many fold on the back end. They come form the same iso 9000 manufacturing lines, and the whole point of ISO9000 is that if you make a turd or a diamond, you make it the same way every time.

Chas said:
I will say this. HGST REALLY hit the jackpot with their 4TB drives!

Those are some fucking CRAZY-LOW failure rates...
Click to expand...

HGST has hit the jackpot with their 2TB drives and 3TB drives, and 4TB drives, and has been looking good in backblaze's stats for years now. It's almsot like they make a consistently good product if you buy their NAS drives.
 
  • Like
Reactions: DocNo
like this
You must log in or register to reply here.
Back
Top