News 71-TiB NAS with twenty-four 4TB drives hasn't had a single drive failure for ten years — owner outlines key approaches to assure HDD longevity

[v2millennium]

High temperature increases the rate of the magnetic surface degradation. Turning off the HDD definitely lowers its temperature.

 

[jeremyj_83]

IF the hdd hit 8000 hrs mark it's time to think of change it. Some disk don't tell right on smart and don't report.
I have one machine with five disk with raid 5... loose my sleep thinking if the motherboard dies. How i will build the raid again.

That is very bad advise. These are enterprise grade drives that are designed to be powered on 24/7. They are warrantied for 5 years and that means the manufacturer is sure they will last longer than that. The actuator of these drives is probably rated for 550TB/ year of writes. Spread that out over 24 drives gives you 10PB/year of writes to the NAS. Even if it is 250TB/year that is still ~5PB/year. Use a RAID array and replace drives when they fail. That makes the most sense.

 

[bit_user]

When you buy enterprise drives, you expect them to last. At work we have a 16 bay NAS filled with 14TB WD Gold HDDs. In the last 5 years of being powered on 24/7 and writes to it all day at random intervals (log backups) we have had 2 drive failures. Therefore this NAS lasting 10 years isn't surprising.

The WD Gold I've seen had a warranty period of 5 years and are specified to handle continuous 24/7 operation. So, it would be pretty bad business for them if like half failed within that time.

Over the next 5 years, you probably won't be as lucky.

 

[kanewolf]

He should consider himself lucky as most failures actually occur during power up where current draw is maximum and this is when things like spindle motors fail.

Yep. Disks are happiest at constant temp with stable power. I worked on a system with 1000s of disks. Power them up once, and leave them alone.

 

[jeremyj_83]

The WD Gold I've seen had a warranty period of 5 years and are specified to handle continuous 24/7 operation. So, it would be pretty bad business for them if like half failed within that time.

Over the next 5 years, you probably won't be as lucky.

Oh I know. We had 2 fail in the warrantied period and get them replaced. That said your chances of any single drive failing is reduced by using RAID. This array will most likely be replaced in the next year as we upgrade our data center.

 

[USAFRet]

Oh I know. We had 2 fail in the warrantied period and get them replaced. That said your chances of any single drive failing is reduced by using RAID. This array will most likely be replaced in the next year as we upgrade our data center.

No, the more drives you have, the chance of a drive failing increases.
The RAID gives you continued uptime, until such time as you can replace the failed device.

But any RAID is not a substitute for a real backup routine.

 

[8086]

First, HGST drives of this vintage were really good.

Second, I had a fileserver with 5x 1 TB WD Black HDDs that also lasted over 10 years with zero unrecoverable sectors on any of the drives. As in his case, I turned it off when not in use. I used it mostly for backups and it probably had a similar number of hours as his.

Finally, I've seen machines at work with HDDs in continuous service for more than 10 years. I think one of them was even running for more than 15 years! So, it's possible to have old hard drives last that long, even when in continuous use!

My old C drive is still going strong in my last two upgrades, with over 2 billion reads/writes and 73000 hours of nearly continuous on time and it shows no signs of bad health (via S.M.A.R.T.) or slowing down.

 

[HideOut]

Tech enthusiast Louwrentius says that turning off the NAS when unused likely helped the 24 4TB hard drives last ten years without any failures.

71-TiB NAS with twenty-four 4TB drives hasn't had a single drive failure for ten years — owner outlines key approaches to assure HDD longevity : Read more

LOL what a missleading headline. Only 6000 hours in 10 years. Thats about 2/3 of a year of run time. So while they may have been instaleld for 10 years, th eya re less than a year old in use wise. It has NOTHING to do with normal NAS operations. You are reporting news like CNN and FOX

 

[Rob1C]

Electricity is fairly expensive in smaller countries, using an online calculator at 10W per drive, he's saved quite a bit over 10 years. There's also the advantage of the air gap / Sneakernet from having them powered off.

 

[Amdlova]

@jeremyj_83 8000 hours on consumer hdd its about 5 years of normal usage.
Enterprise drives are made for controlled environment with chick's and drinks

 

[USAFRet]

Electricity is fairly expensive in smaller countries, using an online calculator at 10W per drive, he's saved quite a bit over 10 years. There's also the advantage of the air gap / Sneakernet from having them powered off.

Considering 4TB Seagate Ironwolf drives:
https://www.seagate.com/www-content/datasheets/pdfs/ironwolf-12tbDS1904-9-1707US-en_US.pdf

Idle consumption is 3.95W
Sleep mode - 0.5W

Savings, yes. But not as much as it might seem.

 

[adamXpeter]

Once upon a time there was a legendary HDD: Quantum Fireball LCT, where LCT meant Low Cost Technology.
These drives were legendary because they lived till 1-3 days after the 1 year warranty expired.
Somehow this kind of HDD ended up in our server which was turned off only three times a year.
They lived 7+ years in that server.

 

[Makaveli]

I was under the impression the start and stop cycles is what stresses a drive.

I run a 20TB NAS at home with Seagate Iron Wolf 3x4TB and 1x8TB and just leave it running haven't been running my drives as long as him but no failures yet. I may try this shut down the nas from time to time. Like when i'm going on vacation etc

ADM seems to have Power scheduling where I can set a power on and off schedule so I may play around with this.



My drives hibernate after 1 hour also.

 

[OneMoreUser]

So only using something occasionally means it last longer, shocker.

Seems to me the same thinking applies to most things except food items, so the same trick to make things last goes for sofas, shoes, tools, books...

 

[AkroZ]

"Given that most hard drives are only rated to last three to five years,"

Hard drives are mainly rated by writes operations (TBW), they considers an average TBW by year for a "normal" usage. If you do not use it, it shoul not deteriorate much as the materials are quite durable.
I have hard drive 30 years old which can boot Windows 95 under 10 secondes (Legacy BIOS).

At work I have replaced a disk with failures that were two year old from a server assembled by another technician, I checked the documentation and it was a desktop consumer drive with the detail for the lifespan calculation, the TBW was equivalent to 14 years of intensive usage on desktop.

For the power up failures, I think it is more due to the power down. A brutal energy cut can cause damages due to the magnetic head not being off the disk and can ray the surface (like a vinyl). And most systems are cutting immediately cooling when powered off which make components overheat.

 

[stevenrix1]

On march the 17th of 2014 I bought a NAS (a synology) with 8 x 4 Tb drives. The NAS has been running flawlessly 24 hours nonstop and I got only 1 drive that failed (a predictive failure). The drives I bought were the eco drives from Hitachi, the coolspin drives. The only maintenance i've done once a year was to remove the dust around my hard-drives. I never updated the firmware (if the drives work then do not update the FW is my rule of thumb).

 

[palladin9479]

Yes, but also your affirmation that most failure happens during power up in no way indicate that is was lucky.
In my experience with servers, I never got a single HD broken during power up.
In the case of the article, he used 24 HD and for lot of years he power down and up the NAS countless times, this is indicative at least of the fact that power cycles are not so dangerous as sometimes claimed.

I think your under some misunderstanding on powerup / down and what "a lot" means.

Firstly going from zero power to full power is absolutely the highest stress period of mechanical HDD's, anything that is worn out is likely to fail right here. This is also the difference between Enterprise / NAS grade HDD's and home / consumer HDD's. The former category spins up / down slower and generally doesn't want to go above ~5400 RPM, this limits the stress on the motors and electronics ensuring a long life. The later will spin up / down quickly and prefers to be at ~7200RPM, this induces more stress but the drives are more responsive. Enterprise / NAS drives also have a quasi idle mode where they go into a low RPM state but not zero, this reduces stress when they are requested to spin back up again. Consumer drives instead go to a full stop 0 RPM parked state.

Now onto the whole power up / down thing, once or twice a day is not a big deal, which is what this guy appears to be doing. 3,650 power cycles isn't much to be concerned about, generally speaking 10,000+ power cycles is were things start getting dicey, and even then it's all down to the quality of the drives.

 

[bit_user]

Considering 4TB Seagate Ironwolf drives:
https://www.seagate.com/www-content/datasheets/pdfs/ironwolf-12tbDS1904-9-1707US-en_US.pdf

Idle consumption is 3.95W
Sleep mode - 0.5W

Savings, yes. But not as much as it might seem.

At my residential rate of electricity and assuming a PSU running at 85% efficiency, running 24 such drives for a year would cost $312.28, not including any added air conditioning costs. Yes, I could turn around and save some $ on heating, if the server weren't located in a room above where the thermostat is located.

IIRC, the idle power of my WD Gold drives is about 7 W, but my current setup is using only four of them.

 

[bit_user]

Hard drives are mainly rated by writes operations (TBW), they considers an average TBW by year for a "normal" usage. If you do not use it, it shoul not deteriorate much as the materials are quite durable.

I think the endurance rating is primarily about wearing out the media. That doesn't mean it's the typical failure mode. I think HDD makers just use the write endurance rating to say that it will fail prematurely, if you use it for a workload more intensive than it was designed to handle. This is partly to warn you not to misuse the drive and partly to protect themselves if you do.

At work I have replaced a disk with failures that were two year old from a server assembled by another technician, I checked the documentation and it was a desktop consumer drive with the detail for the lifespan calculation, the TBW was equivalent to 14 years of intensive usage on desktop.

Sounds like it was swapping heavily. I'd check how much RAM the machine had, and upgrade it if sub-standard. I think 16 GB is pretty standard, these days.

For the power up failures, I think it is more due to the power down. A brutal energy cut can cause damages due to the magnetic head not being off the disk and can ray the surface (like a vinyl). And most systems are cutting immediately cooling when powered off which make components overheat.

Don't they have enough capacitors at least to safely park the head, in the event of sudden power loss?

On a related note, I remember some laptop harddrives incorporated an accelerometer and would self-park, if they detected that they were being dropped. A few years ago, I was fiddling with an accelerometer I got for a project involving my Raspberry Pi. I was amused to see the manufacturer datasheet suggested laptop fall-detection as precisely one of the applications it had been designed for.

 

[bit_user]

I never updated the firmware (if the drives work then do not update the FW is my rule of thumb).

When I get a new HDD or SSD, the first thing I always do is check that it's running the latest firmware. After the drive is in use, I don't mess with the firmware unless there's a known issue that seems worth the risk to address. Two such examples that come to mind:

• https://www.tomshardware.com/news/samsung-980-pro-ssd-failures-firmware-update
• https://www.tomshardware.com/news/samsung-990-pro-ssd-firmware-fix

 

[bit_user]

This is also the difference between Enterprise / NAS grade HDD's and home / consumer HDD's. The former category spins up / down slower and generally doesn't want to go above ~5400 RPM, this limits the stress on the motors and electronics ensuring a long life. The later will spin up / down quickly and prefers to be at ~7200RPM, this induces more stress but the drives are more responsive.

Did you get that backward? Up until a few years ago, it seemed 7200 RPM was pretty common among enterprise and consumer drives would often be 5400 or even variable (WD Green comes to mind). The lowest-speed consumer drives I've seen were 4500 RPM. Enterprise drives used to range up to 10k and even 14k RPM, but those high-speed drives had reduced capacity (2.5" or even 1.8" platters, I think) and were quickly killed off by SSDs.

Enterprise / NAS drives also have a quasi idle mode where they go into a low RPM state but not zero, this reduces stress when they are requested to spin back up again.

I haven't heard of this being done in enterprise drives. Is it for near-line drives? I don't see it mentioned in the product briefs of the latest WD DC Ultrastar or WD Gold drives.

I don't see it mentioned of Seagate's Exos drives, either. They do make a big deal about those drives being low-latency, which isn't consistent with spinning down to lower RPMs.

 

[jeremyj_83]

@jeremyj_83 8000 hours on consumer hdd its about 5 years of normal usage.
Enterprise drives are made for controlled environment with chick's and drinks

This person is using enterprise drives. Even for consumer drives 8000hrs isn't anything. That is less than 1 year of 25/7 usage. For example, my current desktop is 11 years old and has way more powered on hours than that. I've never had issues with my 2TB WD Green drive.

 

[leclod]

He should consider himself lucky as most failures actually occur during power up where current draw is maximum and this is when things like spindle motors fail.

I've got 14000 Start/stop on my oldest drive, so he's got some room.

 

[nitrium]

I have 4 WD Reds : a 1.5TB, a 2.5TB, a 4TB and an 8TB for my NAS that range in age from 4-10 years . Have had no issues whatsoever, and they are on 24/7. The oldest drive (the 1.5TB) gets hammered the hardest too.

 

[Kamen Rider Blade]

At my residential rate of electricity and assuming a PSU running at 85% efficiency, running 24 such drives for a year would cost $312.28, not including any added air conditioning costs. Yes, I could turn around and save some $ on heating, if the server weren't located in a room above where the thermostat is located.

IIRC, the idle power of my WD Gold drives is about 7 W, but my current setup is using only four of them.

Here's a neat way for you to "Permanently Reduce" your AC costs during summers.

Spoiler: Technology Connections - Awnings.

Installing Awnings on any Windows that will face the sun could reduce Solar Heat Gain by 66%-75%.

Apparently it's a piece of simple technology that many of us forgot once "Big AC" industry went into effect.