Hard Disk Life Expectancy and SMART Errors

Since we have quite a long discussion going on here about backups, I thought these links to Backblaze might prove illuminating.

 

Backblaze provides cloud storage, and business and personal backup service. For the past several years they have been posting stories about their experience with disk failures. Since they have over 200,000disks online, they have a large enough set to have valid statistics. Of course, Backblaze uses Enterprise disks as well as Consumer disks

 

LINK: Hard Drive Life Expectancy (backblaze.com)

 

LINK: What SMART Hard Disk Errors Actually Tell Us (backblaze.com)

 

LINK: A Look at How Backblaze Buys Petabytes of Hard Drive Storage

 

Enjoy some reading.

My concern is how a cloud service determines that their disk with my data is about to fail wiping out my data before that happens and moving it to a good disk? How do they handle it if their drive fails wiping out my data? How is the data recovered?

The way I handle it at home with my own local backup is that should the current operational backup disk fail, I still have that data on my desktop. So I can just create a new backup disk. Old data not on the current desktop or current backup disk is stored on a couple of "spare" backup disks. Of course, if my house burns down God forgive, then I'm screwed in a lot of ways worse than my data.

 

If the cloud service loses my data, do they just download the data again from my computer? What if I offloaded some of it already?

The concept of keeping 'synced' copies between drives worries me.

 

How does the system know NOT to replace a good copy with a bad one?

 

ie how does it know the difference between a corrupted file and one I've make legit changes to?

 

Does it have to determine every file is still fully readable before 'updating'?

As far as I know, cloud services, and other big data places, use RAID arrays.

 

That is, with some number, I believe more than one, redundant disk.

 

If one disk dies, the data can be reconstructed from others.

That isn't quite good enough, as it is possible for another to die during

the reconstruction, so it is usual to have an array with two disks

of redundancy, or maybe more.

 

They should also have some form of off-site backup.

How does the system know NOT to replace a good copy with a bad one?

 

@mike Data is written to HDD in sectors. The old standard was a 15-byte header consisting of gap, sync, and address mark, followed by 512 bytes of data, followed by 50-bytes of error correction code for a total of 577-bytes per sector. By 2011 with larger disks, the original sector had been replaced by Advanced Format with the same 15-byte header, followed by 100-bytes of error correction code for a total sector size of 4211 bytes. The old sector had an efficiency 0f 88.7% data to overhead while the Advanced Format had and efficiency of 97.3%

 

Both formats had error correction codes as part of the sector. The header and error correction codes are used by the disk controller, which is part of the disk (on the printed circuit board on the bottom of the disk drive), to ensure the data is valid - not corrupt. The user and the operating system see only the data and not the header and ECC in normal operation.

 

LINK: Disk sector - Wikipedia

 

LINK: Advanced Format - Wikipedia

 

Can error correction codes fail? Yes, they can, but it is a very, very rare event. Nothing is ever 100% assured. Just ask Heisenberg :)

 

If you or your program write "bad (invalid) data to the disk, no error correction code will save you. That is why there are backups, preferably several generations of backups.

My concern is how a cloud service determines that their disk with my data is about to fail wiping out my data before that happens and moving it to a good disk? How do they handle it if their drive fails wiping out my data? How is the data recovered?

 

See my link in the original post about SMART. That is one way they tell and take proactive measures. If the drive fails, they restore from their backups.

See my link in the original post about SMART. That is one way they tell and take proactive measures. If the drive fails, they restore from their backups.

I reread the link and did not find that statement. Probably an oversight on my part. Could you copy from there where it says that and paste it in your post here?

See my link in the original post about SMART. That is one way they tell and take proactive measures. If the drive fails, they restore from their backups.

I missed that. So they have 100% backup? So if I have 6Tb of data, the same data is backed up on two disks totaling 12TB??

Hi Alan,

 

Of course they have backups. They have drives that fail without warning. They have to be able to recover the data, either from backups or by rebuilding the RAID array. I am sure they compress the backup data so your 12 TB may take less the 12 TB in the backup.

 

Many Cloud datacenters keep mirror copies in diverse geographical locations. That caused problems for Amazon and Microsoft when the stored data from EU countries in the US given the difference in data storage laws between the US and the EU.

I reread the link and did not find that statement. Probably an oversight on my part. Could you copy from there where it says that and paste it in your post here?

 

 

"How does understanding the correlation, of lack thereof, of these SMART stats help us? Let’s say, a drive reported a SMART 5 raw value of 10 and SMART 197 raw value of 20. From that we could conclude the drive is deteriorating and should be scheduled for replacement. Whereas, if the same drive had SMART 197 raw value of 5 and a SMART 198 raw value of 20 and no other errors, we might hold off on replacing the drive awaiting more data, such as the frequency of the errors occurring."

Here is a better description of Backblaze's data storage. Note your data does not reside on any given hard drive. It is split like RAID 5.

 

LINK: The Cloud's Software: A Look Inside Backblaze

 

That is for Backblaze. AWS and Azure probably use a different strategy.

 

And here is how you can have your very on version of their storage POD

 

LINK: Open Source Storage Server: 60 Hard Drives 480TB Storage (backblaze.com)

 

BTW, I have never used any of the services, Backblaze, AWS or Azure.

Edited July 31, 2022 by bgelfand

One problem with at least some cloud services is that they don't consider the user to be the customer; rather, the advertisers are their real customers. The people who use the services are the targets. I personally would not trust important data to be kept backed up and safe on these services. It's not that they can't do backups but because they don't care, sometimes due to human error or code bug, backups might not end up being made. Only the owner of the data really cares.

"How does understanding the correlation, of lack thereof, of these SMART stats help us? Let’s say, a drive reported a SMART 5 raw value of 10 and SMART 197 raw value of 20. From that we could conclude the drive is deteriorating and should be scheduled for replacement. Whereas, if the same drive had SMART 197 raw value of 5 and a SMART 198 raw value of 20 and no other errors, we might hold off on replacing the drive awaiting more data, such as the frequency of the errors occurring."

This doesn't say they back up the data twice. All this says is if they suspect a drive will fail, they move the data to another drive beforehand. Where do they explain what happens to data on a drive that fails without notice?

Here is a better description of Backblaze's data storage. Note your data does not reside on any given hard drive. It is split like RAID 5.

 

LINK: The Cloud's Software: A Look Inside Backblaze

 

That is for Backblaze. AWS and Azure probably use a different strategy.

 

And here is how you can have your very on version of their storage POD

 

LINK: Open Source Storage Server: 60 Hard Drives 480TB Storage (backblaze.com)

 

BTW, I have never used any of the services, Backblaze, AWS or Azure.

I don't know what that means or what that has to do with them backing up data.

 

Can you find and copy from that article(s) where it says they back up the data twice so customers won't lose their data rather than asking us to read two more articles? I think you ought to back up your claim about this. (no pun intended).

I don't know what that means or what that has to do with them backing up data.

 

Can you find and copy from that article(s) where it says they back up the data twice so customers won't lose their data rather than asking us to read two more articles? I think you ought to back up your claim about this. (no pun intended).

A RAID array, except for the lowest level, is intended to keep data safe when a drive or drives they are on fail. A backup, mirror, is the simplest way. Other levels distribute data over multiple disks and will reconstruct any data that was on a disk that failed. That indeed is a good way to keep data safe when disks fail.

But yes, mirroring a RAID array on another RAID array ads another level.

A RAID array, except for the lowest level, is intended to keep data safe when a drive or drives they are on fail. A backup, mirror, is the simplest way. Other levels distribute data over multiple disks and will reconstruct any data that was on a disk that failed. That indeed is a good way to keep data safe when disks fail.

But yes, mirroring a RAID array on another RAID array ads another level.

Does Backblaze use this method of RAID? How does RAID work to prevent the loss of data?

Does Backblaze use this method of RAID? How does RAID work to prevent the loss of data?

This article Backblaze Storage Pod claims RAID6 is being used - not sure whether that's still the case. From the little I understand of all the talk about pods and vaults and the various pieces of software is that there appears to be a high level of redundancy giving the option to restore missing data (from disk failure, for example) to a very high degree. Doesn't answer the question though if the data of one pod or vault (or an entire data center) is mirrored somewhere else or if one completely relies on re-creating the data on failed drives. I believe it is the latter though hope that there exists at least one physical copy somewhere else (though I doubt that's the case for simple cost reasons as well as the stated fact that their data durability is 99.99999% using their method of storing and reconstructing if needed).

 

I toyed with the idea of creating a RAID array but didn't go that route. Need for constant maintenance and reliance on a piece of software/hardware were my main reasons not to pursue RAID. I prefer to just do 1:1 copies (no compression, no need for any piece of software to read the backup later) to at least two different drives for all my backup needs. Automatically twice a day for my general files and started manually for the images whenever I come back from a photo outing. Periodically, I update the off-site backups.

 

My internet connection has always been too slow and/or unreliable/unstable to seriously consider cloud backup. Even now with a relatively fast connection, it would take 20-30 seconds for each D850/A7RIII RAW file to upload at the highest upload speed available.

 

I recall that at one startup company we decided to build our own RAID backup system when we realized that the data we produced per day would take much longer to back up to the cloud. We ended up going with two RAID servers and tape backup for long-term off-site storage.

Edited July 31, 2022 by Dieter Schaefer

Any decent RAID array is going to include a parity check between mirrors, and a non-zero parity bit XOR will surely raise an alarm?

Therefore any failing disk, PSU glitch or whatever, ought to be caught (and addressed) early, before any permanent data loss results.

 

My concern with 'Cloud' storage wouldn't be data loss, but data retention; together with a sharp hike in the price of renting your online storage. It's not like the cost of personal ownership of multi Terabytes of HDD real estate is expensive these days. Nor particularly bulky.

My main concern with cloud storage is that they form a populair target for hackers especially now in with the Russia in Ukraine situation :

 

How cloud services become weapons in Russia-Ukraine cyber conflict | ZDNet

My main concern with cloud storage is that they form a populair target for hackers especially now in with the Russia in Ukraine situation :

... Plus (1) we pass our own ineptitudes to other people who may have other negative issues and circumstances. (I can tell you scary incidents that were never spread outside of our IT unit when I worked there). (2) Upload time for huge files may be intolerable.

 

PS It is not a bad option, provided we don't put all the eggs in one basket.

Edited July 31, 2022 by Mary Doo

... Plus (1) we pass our own ineptitudes to other people who may have other negative issues and circumstances. (I can tell you scary incidents that were never spread outside of our IT unit when I worked there). (2) Upload time for huge files may be intolerable.

 

PS It is not a bad option, provided we don't put all the eggs in one basket.

 

I have worked in IT from 1982 on as a software engineer on verry large systems as wel on middleware and networked PC level platforms, so i think i have seen my fair share of "scary incidents"...:);):cool:

Why has this thread staggered along only in the Nikon forum? It's surely not a brand-specific issue! MIA mods yet again? Sad...

Do you not have better things to do?

Why has this thread staggered along only in the Nikon forum? It's surely not a brand-specific issue! MIA mods yet again? Sad...

Everyone is free to tour all the different forums. No issue there.

 

If Nikon RAW NEFs are a particular issue, then yes, I want to find it here. We all have individual IT based issues.

 

From a specific forum, you need people who's opinion you can trust. I'm not going to ask the Spud-U-Like forum IT questions, not 'cos I don't necessarily believe the info they might produce, but I just don't know them.

 

staggered along

The only thing staggering along here is you.

 

Get it?