HDD reliability

[SameSelf]

This implies that it is wrong to back up your drives, period.

Even if using a USB hub, with two slots for two drives, one with your data and the other with a backup of that data, it is still better than only using one drive to store your data. Even that link supports backups.

Sure there are problems, and still lots can go wrong. But your chances are still better when you have your data in more than one place. That was my point.

If you had read that more carefully, you would have understood that it doesn't matter how many drives are plugged into your 'puter, data drive, back up drives, RAID 5/6. Once silent data corruption starts to occur, it will cascade across every copy undetected thereby infecting the "back-ups", "back-ups of back-ups", and so on. But trying to explain this is challenging. It is easier to simply ask:

What are you doing to guard against silent data corruption?

[PuzZLeR]

This implies that it is wrong to back up your drives, period.

Even if using a USB hub, with two slots for two drives, one with your data and the other with a backup of that data, it is still better than only using one drive to store your data. Even that link supports backups.

Sure there are problems, and still lots can go wrong. But your chances are still better when you have your data in more than one place. That was my point.

If you had read that more carefully, you would have understood that it doesn't matter how many drives are plugged into your 'puter, data drive, back up drives, RAID 5/6. Once silent data corruption starts to occur, it will cascade across every copy undetected thereby infecting the "back-ups", "back-ups of back-ups", and so on. But trying to explain this is challenging. It is easier to simply ask:

What are you doing to guard against silent data corruption?

Sure this sucks, and I do appreciate the heads up.

Help me understand. Are you implying that the actual act of placing your data in more than one place creates silent data corruption, or compounds it, and infects the original?

[bendixG15]

Don Quixote has so many windmills to attack............

[PuzZLeR]

Don Quixote has so many windmills to attack............

Which is the imaginary giant in this case?

[SameSelf]

Help me understand. Are you implying that the actual act of placing your data in more than one place creates silent data corruption, or compounds it, and infects the original?

No I am not implying that. Here is how a typical backup plan works.

You have a data drive (internal, external, doesn't matter). You have a second drive that is used to back up the data on the data drive (again, internal, external, doesn't matter). You run some sort of back up software program. That program runs at some frequency making incremental backups because it is too costly to re-write an entire hard drive's contents just because a handful of new files appeared or were changed. So the backup software looks for the files that were changed or new. If one of those files has a bit flip (silent data corruption) the software reads that as a change and overwrites the backup file with the bit flipped version. Now both the back up drive and the data drive have a corrupt version of the file and the original uncorrupted version is gone forever unless you just happened to have it saved somewhere else. In fact, there is no intrinsic way for any program to determine if a file is corrupt unless it has a checksum validation as part of the indexing. The only file system even remotely capable of doing that is ZFS.

Hope that helps.

[PuzZLeR]

Hi again. Thanks for the explanation. And I get it. And you have encouraged me to run checksums more frequently, and I have indeed noticed this problem (ex: corrupted some Excel and JPG files) when running programs like ActiveSync in the past (so since, I do it manually, the long way).

But the point being is that this is a problem with data and storage, files, file systems, etc., not with backing up - unless you want to include a problem with backup software as a problem with backing up.

Yes, backing up compounds a problem, but I understand it as a problem that was already there before, not a problem that backing up creates.

If one of those files has a bit flip (silent data corruption) the software reads that as a change and overwrites the backup file with the bit flipped version. Now both the back up drive and the data drive have a corrupt version of the file and the original uncorrupted version is gone forever unless you just happened to have it saved somewhere else.

The bit flip/error in your explanation happened before the backup, and in the end actually encourages a backup regardless with the original. I understand how the backup is spreading the problem, but how is it creating it in the first place?

[smitbret]

I have been following the reports from BackBlaze since they started publishing them and here are the takeaways that I get:

#1 - 3TB Seagates have no business in my server
#2 - 1.5TB Seagates have no business in my server
#3 - If I am shopping for a new HDD and the Hitachi/HGST is within $10 of everyone else I grab the Hitachi

Otherwise, price wins with a small nod to warranty if all else is equal.

[Groucho2004]

"Those who backup their data regularly are less likely to have HDD crashes."

That statement makes no sense.
You are basically saying that the drive being backed up has a conscience and is aware that a backup exists and therefore decides not to crash.

At face value you have a point. But it's something that makes you think.

I'm talking about confidence. You can't see it, or hear it, or smell it, or touch it, or even "feel" it in a conventional way. But you can feel it in spirit, and sense it in faith. You know it's there, and creates positive results. And I know you believe in the power it emits.

Seriously? Do I get less crashes if I sacrifice a goat with each backup?

When you back up your drives you put less strain overall per unit on your overall GBs. This alone is enough to extend life and minimize HDD crashes. :-p

Again, makes no sense. In fact the opposite is true. Every time you run a backup you put strain on both the source and the backup drive.

Don't get me wrong, I'm all for redundancy and I backup all my data regularly to at least two different locations. You just seem to be getting your facts from an alternate universe where the laws of physics are different.

[SameSelf]

The bit flip/error in your explanation happened before the backup, and in the end actually encourages a backup regardless with the original. I understand how the backup is spreading the problem, but how is it creating it in the first place?

Sorry if my explanation is confusing. The cause of the bit flip error is simply a manifestation of the second law of thermodynamics or entropy. A more concrete example might be: the sun sends out squillions of neutrinos every second and virtually all of them pass through the earth, us, and hard drive platters unimpeded, but if just one happens to interact with an electron on a platter on that neutrino's way to outer space then a 1 might become a 0. Researchers aren't completely certain all the causes. But it probably happens more often than we realize. That is what the geniuses at CERN discovered. They found it happens more often than hard drive mfr's claims, no surprise there I suppose and a major downer for the people who pay attention to these things. Can hard drive mfr's do anything? They sure haven't given any indication they can.

As for the backup software being the problem, that is not entirely accurate. The problem is that backup software and even RAID solutions, both hardware and software, have no way of knowing that a bit flip is not a file change initiated by the user. They are agnostic. Data redundancy implemented in the form of incremental backup software or RAID is UNABLE to guard against silent bit corruption. I don't make the rules, so don't kill the messenger. I have said already the only solution is a filesystem that implements checksums, COW, etc and ZFS is the only free candidate I am aware of.

Now you might think I am overblowing this problem. Sadly I don't think I am. This is a video forum. My situation is probably similar to most. I have over 1 TB of master video files that average over 10 GB per master. So we are not talking about a handful of JPGs. It would just take one bit flip to ruin an hour of video.

EDIT: This is the best online resource I have found describing the problem. It is technical but worth diving into.

[Clockwork]

We've survived thousands of years as a civilisation without being burdened with "flipped bits".
Let's try and keep things in perspective.

[SameSelf]

??? Try telling that to the researchers at CERN....Paper

[bendixG15]

Don Quixote has so many windmills to attack............

Stop adding more damn windmills.

[hello_hello]

Still don't buy the reasons unless Backiblaze doesn't mind just tossing out drives. Surely the warranty on consumer drives doesn't cover commercial applications. At least of an Enterprise drive fails, they could get it replaced under warranty.

Would the drive manufacturer know what application a drive was used in? You buy a drive, it dies within the warranty period, it gets returned for a replacement.

Spinning rust is a commodity business. It is cheaper for mfg's to replace your ailing drive under warranty than to build a drive that is meant to survive the zombie apocalypse. They engineer the reliability to stay profitable. WD even provides a free diagnostic tool to tell you when the drive has gone bad. What else should that tell you?

It could tell you manufactures are trying to minimise the return of healthy drives. Most manufacturers have a similar tool. Seagate tell you to run their's before returning a drive because they require the error code it supplies on their RMA form (although they do have a "can't run the diagnostic tool" code you can use instead). Manufacturers wouldn't check returned drives individually. They probably just replace returned drives and the returned drives would be refurbished.

I installed hard drive 'coolers' on the most frequently used desktop computers. With the cases open, I could really feel a difference in the temperature of the hdd as opposed to not using a cooler at all. My theory being (among other things), that heat can contribute to hard drive failure and mechanical wear.

When Google released the results of their hard drive study several years ago, they concluded hard drive temperature had almost no correlation to longevity. In fact if I remember correctly they found cooler running drives had very slightly higher mortality rates in the first three years.

I think it is well known that modern drives are not as reliable as drives from 10 or 20 years ago.

My experience is pretty much the opposite. Back in the days of capacities under 100 GBs, most drives I bought were dead within a couple of years, however I've still got 320GB/500GB WD/Samsung drives going strong. Last time I checked a couple, their total running time was over 4.5 years, and they're just standard desktop drives.
The only manufacturer I've had "bad luck" with in quite a while is Seagate. I bought a couple of 500GB Barracuda drives that turned out to be a problem model and I had to return them at least twice. Having said that, I'm pretty sure the oldest surviving hard drive in this house is a 20GB Seagate. It lives in a small form factor Pentium 3 PC that's still used a couple of hours per day on average (simple games).

[PuzZLeR]

The bit flip/error in your explanation happened before the backup, and in the end actually encourages a backup regardless with the original. I understand how the backup is spreading the problem, but how is it creating it in the first place?

Sorry if my explanation is confusing. The cause of the bit flip error is simply a manifestation of the second law of thermodynamics or entropy. A more concrete example might be: the sun sends out squillions of neutrinos every second and virtually all of them pass through the earth, us, and hard drive platters unimpeded, but if just one happens to interact with an electron on a platter on that neutrino's way to outer space then a 1 might become a 0. Researchers aren't completely certain all the causes. But it probably happens more often than we realize. That is what the geniuses at CERN discovered. They found it happens more often than hard drive mfr's claims, no surprise there I suppose and a major downer for the people who pay attention to these things. Can hard drive mfr's do anything? They sure haven't given any indication they can.

As for the backup software being the problem, that is not entirely accurate. The problem is that backup software and even RAID solutions, both hardware and software, have no way of knowing that a bit flip is not a file change initiated by the user. They are agnostic. Data redundancy implemented in the form of incremental backup software or RAID is UNABLE to guard against silent bit corruption. I don't make the rules, so don't kill the messenger. I have said already the only solution is a filesystem that implements checksums, COW, etc and ZFS is the only free candidate I am aware of.

Now you might think I am overblowing this problem. Sadly I don't think I am. This is a video forum. My situation is probably similar to most. I have over 1 TB of master video files that average over 10 GB per master. So we are not talking about a handful of JPGs. It would just take one bit flip to ruin an hour of video.

EDIT: This is the best online resource I have found describing the problem. It is technical but worth diving into.

Your explanations are not confusing when describing the problem, and I don't believe you're overblowing it when it may be of interest to the users of this Forum.

I'm just stating that this exchange began when you suggested that backing up is the problem. Backing up spreads the problem, yes, but I can't see how it creates it in the first place. That was my point.

Backing up is the messanger only, not the culprit.

Also, backing up semi-useful data can still be better than not backing up anything at all.

[PuzZLeR]

"Those who backup their data regularly are less likely to have HDD crashes."

That statement makes no sense.
You are basically saying that the drive being backed up has a conscience and is aware that a backup exists and therefore decides not to crash.

At face value you have a point. But it's something that makes you think.

I'm talking about confidence. You can't see it, or hear it, or smell it, or touch it, or even "feel" it in a conventional way. But you can feel it in spirit, and sense it in faith. You know it's there, and creates positive results. And I know you believe in the power it emits.

Seriously? Do I get less crashes if I sacrifice a goat with each backup?

Hint exposed: My full post there (with no edits) said it's a saying - I did explicitly state this - it doesn't need to be taken literally or logically. Its objective is in stimulating thought, not necessarily in direct application (which is true of many sayings). And it is indeed a saying in the industry (not mine) and I did post a clear source of this in a subsequent post.

Unfortunately, a tinge of humor (or even an attempt at philosophy) seems to execute poorly in tech forums, and at face value they read like bad facts.

But somehow, in a metaphysical, or even religious, way, I do believe in the saying myself. I do say, and did say in the full post there too, "believe" - not that it's true. If one here believes in the Great Kumquat as their Diety, who am I to challenge this?

When you back up your drives you put less strain overall per unit on your overall GBs. This alone is enough to extend life and minimize HDD crashes. :-p

Again, makes no sense. In fact the opposite is true. Every time you run a backup you put strain on both the source and the backup drive.

Don't get me wrong, I'm all for redundancy and I backup all my data regularly to at least two different locations. You just seem to be getting your facts from an alternate universe where the laws of physics are different.

Again, this was merely an example, and can be construed in many ways.

I was stating strain "per unit" on all drives, NOT strain on less drives.

It depends on how you back up, particularly if you use one+ backup per one source. If you use one drive to encode, edit, capture, rewrite/save a Word/Excel file a million times, and all other sorts of plumbing, then run a simple copy to a backup when done, well, this is much less strain on the backup than the source drive.

This is a strain on two+ drives, yes, but not as much "per unit" on two+ drives collectively, and less so collectively per unit when copying to more and more drives. And, in the long run, your data is preserved on less strain per unit mathematically with backups.

[SameSelf]

Still don't buy the reasons unless Backiblaze doesn't mind just tossing out drives. Surely the warranty on consumer drives doesn't cover commercial applications. At least of an Enterprise drive fails, they could get it replaced under warranty.

Would the drive manufacturer know what application a drive was used in? You buy a drive, it dies within the warranty period, it gets returned for a replacement.

Spinning rust is a commodity business. It is cheaper for mfg's to replace your ailing drive under warranty than to build a drive that is meant to survive the zombie apocalypse. They engineer the reliability to stay profitable. WD even provides a free diagnostic tool to tell you when the drive has gone bad. What else should that tell you?

It could tell you manufactures are trying to minimise the return of healthy drives. Most manufacturers have a similar tool. Seagate tell you to run their's before returning a drive because they require the error code it supplies on their RMA form (although they do have a "can't run the diagnostic tool" code you can use instead). Manufacturers wouldn't check returned drives individually. They probably just replace returned drives and the returned drives would be refurbished.

I installed hard drive 'coolers' on the most frequently used desktop computers. With the cases open, I could really feel a difference in the temperature of the hdd as opposed to not using a cooler at all. My theory being (among other things), that heat can contribute to hard drive failure and mechanical wear.

When Google released the results of their hard drive study several years ago, they concluded hard drive temperature had almost no correlation to longevity. In fact if I remember correctly they found cooler running drives had very slightly higher mortality rates in the first three years.

I think it is well known that modern drives are not as reliable as drives from 10 or 20 years ago.

My experience is pretty much the opposite. Back in the days of capacities under 100 GBs, most drives I bought were dead within a couple of years, however I've still got 320GB/500GB WD/Samsung drives going strong. Last time I checked a couple, their total running time was over 4.5 years, and they're just standard desktop drives.
The only manufacturer I've had "bad luck" with in quite a while is Seagate. I bought a couple of 500GB Barracuda drives that turned out to be a problem model and I had to return them at least twice. Having said that, I'm pretty sure the oldest surviving hard drive in this house is a 20GB Seagate. It lives in a small form factor Pentium 3 PC that's still used a couple of hours per day on average (simple games).

You need to edit your post. That second of the quotes supposedly by me, aka Sameself, is not mine but the OPs. As much as I would love to take credit for everything posted on this thread, I humbly acknowledge that one is not mine.

[hello_hello]

You need to edit your post. That second of the quotes supposedly by me, aka Sameself, is not mine but the OPs. As much as I would love to take credit for everything posted on this thread, I humbly acknowledge that one is not mine.

Fixed. I quoted a post that contained quoted quotes within quotes and I guess I deleted the wrong ones. The quote should point to the post I quoted now, but don't quote me on that.....

[ron spencer]

All this talk of bit rot is great and informative. But here is the catch. If one has running a ZFS RAID 1 or RAID whatever, great. You have a nice robust backup and the filesystem will take care of itself and you can relax.

But even if one sets it up, pretty much everyone will be backing up files from their local (likely Windows) PC to the ZFS share. If you get an issue on the Windows machine (say bit rot) and you are using software to automatically backup your files, those damaged files will be sent to the ZFS share and you can loose your data. How can this be overcome? Unless the ZFS system becomes the system you work and do other backups from.

[SameSelf]

All this talk of bit rot is great and informative. But here is the catch. If one has running a ZFS RAID 1 or RAID whatever, great. You have a nice robust backup and the filesystem will take care of itself and you can relax.

But even if one sets it up, pretty much everyone will be backing up files from their local (likely Windows) PC to the ZFS share. If you get an issue on the Windows machine (say bit rot) and you are using software to automatically backup your files, those damaged files will be sent to the ZFS share and you can loose your data. How can this be overcome? Unless the ZFS system becomes the system you work and do other backups from.

Missed your post, so sorry for the delayed response.

I would not deploy a FreeNAS + ZFS box as a "backup" solution providing redundancy for ntfs or hfs vols. FreeNAS + ZFS should be the primary place where data sits. FreeNAS + ZFS + ECC ram provide all the modern hardening (checksumming, parity, COW, etc.) that the home user can expect. Then the only missing piece is a cold storage backup solution to suitable WORM media like LTO or dl BD-R.

More to the point: TB+ ssd's are mostly within the reach of consumers. I am planning to completely eliminate spinning rust from my PC. I will have two drives in my computer: the OS ssd and a work drive ssd for my video projects. As soon as a project is complete, move the finished masters and project files to the FreeNAS box and burn a BD-R with the checksums.

EDIT: I should add that cold storage is crucial to protect oneself from ransomware.

EDIT: Oh, and one final edit. With FreeNAS + ZFS you eliminate the need for some crappy incremental backup software. You know the names.

[ron spencer]

All this talk of bit rot is great and informative. But here is the catch. If one has running a ZFS RAID 1 or RAID whatever, great. You have a nice robust backup and the filesystem will take care of itself and you can relax.

But even if one sets it up, pretty much everyone will be backing up files from their local (likely Windows) PC to the ZFS share. If you get an issue on the Windows machine (say bit rot) and you are using software to automatically backup your files, those damaged files will be sent to the ZFS share and you can loose your data. How can this be overcome? Unless the ZFS system becomes the system you work and do other backups from.

Missed your post, so sorry for the delayed response.

I would not deploy a FreeNAS + ZFS box as a "backup" solution providing redundancy for ntfs or hfs vols. FreeNAS + ZFS should be the primary place where data sits. FreeNAS + ZFS + ECC ram provide all the modern hardening (checksumming, parity, COW, etc.) that the home user can expect. Then the only missing piece is a cold storage backup solution to suitable WORM media like LTO or dl BD-R.

More to the point: TB+ ssd's are mostly within the reach of consumers. I am planning to completely eliminate spinning rust from my PC. I will have two drives in my computer: the OS ssd and a work drive ssd for my video projects. As soon as a project is complete, move the finished masters and project files to the FreeNAS box and burn a BD-R with the checksums.

EDIT: I should add that cold storage is crucial to protect oneself from ransomware.

EDIT: Oh, and one final edit. With FreeNAS + ZFS you eliminate the need for some crappy incremental backup software. You know the names.

You are right about the ZFS box being the primary spot to store (with additional backups to local NTFS or whatever hard drives coming FROM the ZFS box). But the average user will not do this. I have one setup and have yet to set it up that way LOL. I suppose I should, even though I already have multiple backups....just not sure how anal I should be. Probably more I suspect.

99.99999% of users will not set this up right. How prevalent is bit rot anyway? Are there any hard core stats on this?

[ron spencer]

Do you need to scrub to take advantage of ZFS fully?

[SameSelf]

You are right about the ZFS box being the primary spot to store (with additional backups to local NTFS or whatever hard drives coming FROM the ZFS box). But the average user will not do this. I have one setup and have yet to set it up that way LOL. I suppose I should, even though I already have multiple backups....just not sure how anal I should be. Probably more I suspect.

99.99999% of users will not set this up right. How prevalent is bit rot anyway? Are there any hard core stats on this?

I agree. Rolling your own is not for the average home user and a pain in the derriere for the computer savvy vs an off-the-shelf home appliance. But,

1) This is a video forum. If you are like me, you deal with massive amounts of data unlike the average home user. My personal video library is over 1 TB now. And now that I shoot primarily in ProRes HQ at 220 Mbps, it grows to the tune of about 100+ GB per month. No way of hosting that on the cloud. So I am in NAS territory whether I like it or not.

2) Sadly, the home appliances, like Synology (which is a great product, those guys are doing some good things), don't get you fully there. Enterprise grade is not an option for the home user, and arguably not even the best option. But, thanks to FreeNAS + ZFS, rolling your own is a not just a good option but a great option. Here is a really great step-by-step guide (not mine) using FreeNAS + ZFS (just ignore the hardware/build steps, but those are good too). So I think setting it up right is definitely in the grasp of anyone who has a modicum of skill.

How prevalent is bit rot? Hard to say, and a function of how big your data store is. But I am feeling more and more the need to harden my hardware. It is annoying when a 4+ hour encode has a problem, and I have to do it all over again.

EDIT: You don't need additional backups to ntfs vols FROM ZFS. That is what ZFS is for. At that point, the only thing missing in your data retention plan is diversity. So ditch the spinning rust and think optical media or tape.

[vhkuho]

which brand of desktop hard drive is more reliable..... Seagate or Western Digital ? Opinions, please

[SameSelf]

"more reliable" is a very vague concept when one considers that all brands experience failures. There is no Holy Grail in the hard drive world. In fact, there is a race to the bottom taking place right now. IOW, does it matter to you if a drive catastrophically fails in 6 months or 4 years? Presumably you cared just as much about the data at 6 months as you would at 4 years.

My advice:

1. Buy the drive with the longest warranty if money is not too much of a concern
2. Back up your data to something else besides a hard drive.

And, welcome to the forums

[PuzZLeR]

Honestly, if I have a drive crash after 5 months, or after 4 years, it makes very little difference to me when I have the data backed up. If the drive lasts long, great. If the drive dies prematurely, then I have a warranty for this. Given this, even if I was on a budget, I wouldn't care if a drive dies prematurely.

So, IMO, it makes little difference what brand you use when your data is safe.

I'm sure someone here may disagree, but from my experience I've had drives I've babied and crashed, and others I've abused badly (encodes, captures, etc), and even traveled with, are still around many years later, and the other way around, and with all brands I've used. I have found no rhyme or reason to this (but then again, after my 100TB+ of usage over the years over many drives, I'd say Seagates were weakest IMO).

Preservation of my data is much more important to me than fussing over potential variables that create drive failures. I just make it a habit to just buy a couple of new drives every so often. Trust me, there's no better relief after a drive crash to see that you still have your data safe elsewhere - everything else means nothing IMO.

[smitbret]

which brand of desktop hard drive is more reliable..... Seagate or Western Digital ? Opinions, please

I will give you a straight answer.

As a brand, Western Digital seems more reliable than Seagate.

I buy Seagate because they are cheaper and are 99% as realiable. I also buy Toshiba when they are almost the same price as Seagate. If I can get an HGST within $10-20 then I go HGST because they seem to be more reliable than Seagate or WD.

Generally, though, buying a HDD because of the brand is short sighted and stupid.

[ron spencer]

which brand of desktop hard drive is more reliable..... Seagate or Western Digital ? Opinions, please

Whichever is on sale...doesn't matter brand because you have multiple backups right??

[fritzi93]

I get Seagates for the same reason BackBlaze does: price. And I have multiple backups always.

I had two of the infamous Seagate 3 TB drives, and one just died after 3 years hard usage. [shrugs]

[SameSelf]

I prefer buying drives with a long (5 year) warranty cuz the mfr's will replace them in year 4 at no cost versus going for the bottom of the barrel bargain basement drive that only comes with a 2 year warranty. The spread is around $20/TB, not bad for an extended warranty.

I should also add that I stopped purchasing hdd's online due to too many DOA. Shipping abuses a drive. But I am fortunate because I have a Microcenter in my town.

[PuzZLeR]

I should also add that I stopped purchasing hdd's online due to too many DOA. Shipping abuses a drive. But I am fortunate because I have a Microcenter in my town.

Missed mentioning this because they weren't drives that have failed after usage, but yeah, should have included this in my post as a big drive killer - shipping.

I get about almost 10%-15% DOA - very frustrating. Yes, I always get them replaced, but it's a real pain trying to make them work, finding out sometimes after hours of testing or overnight failed formats they don't work, sometimes even losing a complete backup, etc. Sometimes it's not obvious right away that they're dead, because they're "only defective" at first.

Then, requesting an RMA, printing the label, repacking it, driving to the nearest courier station, waiting again for the replacement, testing again, etc, etc, yaddy-yah - It feels like a part time job to do all this work!