Long-Term Image File Storage Choices

It is surprising the lack of support for the M disk storage option. All the other types of digital storage are acknowledged to last no more -at a stretch- of 20-30 years. ( I still have a archival CD burned in the 1980s that is readable) Yet testing of M disks will hold data for past 100 years. Yet, everyone discounts this option. It comes down to what is the purpose of storage? If it is truly 'archival' and expected to be accessible in 50-100 years of basement museum type storage, then HHD, SSD, Tape, is out. M disks will -at this stage of technology- be the only option. If the data is just expected to be needed to be there for 5-10 years, HD is better than SSD or flash. For really non critical quickly changing data that is continually used (refreshed) SSD is anyones best choice because of ease of use and speed.
Well, not everyone. :-)

https://www.dpreview.com/forums/post/62256529
 
There is no archival digital media.
Absolutely correct
The only long-term (and more important, short-term!) solution is:

1) Two or more separate backups;

2) Two or more locations;

3) Refresh the backups regularly (yearly or so).
4) If you are really thinking about archiving (20/50/100s of years), be prepared to move your backup from your current media to whatever new media becomes the "new standard". Archiving is a process, not a one-time deal.

I've been through migration from floppy disks to mag tape to CD/DVD's but the current favorite for price/performance is HDDs.
Yes; my redundancy today is HDD on-site + HDD off-site. At some point I'm going to move to HDD on-site + cloud. And I have no doubt that the HDD on-site branch may evolve to a different medium (although magnetic HDD has surpassed everyone's predictions on long-term density and cost improvements).

By the way -- each backup branch needs to provide you with file-level restore as well as system (OS + apps) restore. Many of the cloud solutions only offer data file storage, which is not complete enough for me.
 
"It is surprising the lack of support for the M disk storage option. "

Which makes it less attractive, there is no merit in having a backup with perfect data if it isn't easy to get the hardware to read it.

I often wonder how the reported life-span is arrived at.
 
Lots of good backup advice has been given. However, one that is often forgotten is printing. I would suggest printing your best and most important photos and storing them as you would have pre-digital. Often after you pass on, the prints are all that your family will keep.
 
I'd like to see a fast tape. Would be something if faster than the disk arrays from Barracuda for instance.
Since we were talking large business, the newest revisions of LTO/Ultrium can do 300-360MB/s uncompressed. Given that you typically see these in robotic libraries with 4-8 drives, each with their own dedicated fibre channel or SAS link, that is 1.4-2.8TB/s. If you enable compression, you can go faster, but it's unpredictable. At those speeds, the tape is usually not even the bottleneck.

Not reasonable for home or small business use, obviously.
That sounds impressive. I wonder how they get the speed? I saw a demo from Cheyenne years ago where they took4 or 5 tape drives and combined them in a RAID array to get very fast speeds from slow tape drives. But pretty impractical and it went nowhere that I know of.
 
Optical and Magnetic media easily suffer from bit degradation, a.k.a bit rot, a.k.a. silent data corruption. Use ZFS files to guarantee your data remains exactly the same as you copied and verified. Otherwise some of your images could soon be short only showing a fraction of the total, others will have huge colorful streaks, big white blotches, sprinkles, and snow.
for my workflow, in which part should I try to implement ZFS ?

1. in my pc (dual boot win10/ Fedora Linux)
Forget about ZFS on Windows, I don't believe it is well supported. Debian based Linux distros (i.e. Ubuntu and Mint) make it easy to install the zfs-fuse package so you can read ZFS filesystems. You need a kernel extension to run ZFS in system space, which is probably the way you'd want to do it.
2. NAS (synology)
As far as I know, Synology and QNAP do not support RAID-Z.

Possibly a very power-efficient server could be used to make a home-grown RAID-Z server. Maybe with ARM processor, because Intel is either power hungry or slow, take your pick.
3. external HDDs used as off-site backups
I use NTFS for this because it is more universal than any other good filesystem (FAT32 alas does not qualify as good).
what about other filesystems as Btrfs ?
Btrfs is supposed to be more durable than Ext4, however I think ZFS is a better option. On Linux Mint, Btrfs is a choice for Timeshift, instead of Rsync.

Good questions.
 
Last edited:
Lots of good backup advice has been given. However, one that is often forgotten is printing. I would suggest printing your best and most important photos and storing them as you would have pre-digital. Often after you pass on, the prints are all that your family will keep.
Good advice!

Sometimes PC Talk forum posters seem to think their photo collections (or whatever) deserve a lot of effort, such as replacing HDD sets every 5 years.

Wow, my heirs will never do that.

But I'm sure they will look at printed photos stored in boxes after I'm dead.
 
Last edited:
You should consider the Internet... Your life will be infinetely easier and troublefree.

Then the Cloudprovider (select a reputable one :-)) will have to handle technology shifts, dataloss prevention, multiple copies and so on... And the good ones does that on a level that FAR FAR superseedes any efforts you can do on your own. And you have the most simple interface to get your images again if you ever need them.

Just sayin’ - Life is to short to start handling hardware for your requirements.
 
RAIT, not RAID, surely?
 
You should consider the Internet... Your life will be infinetely easier and troublefree.

Then the Cloudprovider (select a reputable one :-)) will have to handle technology shifts, dataloss prevention, multiple copies and so on... And the good ones does that on a level that FAR FAR superseedes any efforts you can do on your own. And you have the most simple interface to get your images again if you ever need them.

Just sayin’ - Life is to short to start handling hardware for your requirements.
Yep, that's what I use Backblaze for. To me, the #1 most important thing is that the backups actually get made, get rotated, get taken offsite, etc... How many of you have set up an elaborate backup strategy and then parts of it just don't really happen that often? To me, that was the primary flaw in my previous backup strategies. Life gets busy real easy and some parts of the strategy just don't get executed.

Using Backblaze (there are other similar services too), all of that is completely automatic and I only pay $50/yr for unlimited backup storage. When I load some new photos on my PC, they are backed up to the cloud automatically in the next hour.

I also have an internal backup drive which I run a nightly backup too using Acronis True Image, but honestly I probably don't need it.

But, the best thing about this backup strategy is that I get redundant backups, versioned backups, offsite storage and it's all entirely automatic so it never fails to get done.
 
But I'm sure they will look at printed photos stored in boxes after I'm dead.
You might want to step over to the Printers and Printing forum and examine the posts about longevity. There is no "permanent" solution to backup. I've got my Dad's Kodachromes from the late 1930s and some of them faded (most didn't).

Long ago I was managing an IBM 34 system and the building took a weekend to repair its electrics, so we shut down a computer that had been continuously on for several years. When we booted it up on Monday the HD failed. Then we discovered that the backup 8" floppies (!) were busted because we had backed up too much data for them. Basically our system and records were un-recoverable. Took us two months to rebuild the books from paper copies.

I think 99% of us learn from major mistakes like that. But in the spirit of the thread, here is what I do:
  1. Outboard SATA station with two HDs, one in the station and one at the bank. The way tech is going, I suspect that UBS3 HDs are just as good these days.
  2. Backup data (D) and image the OS (C) every couple of weeks (do it every day if you feel like it).
  3. Every month do a backup on the in-house HD and swap it with the one at the bank. Fire up the incoming HD as soon as possible and do a backup.
This way I have my stuff in three places. The chance of all three drives going down at the same time is, barring nuclear attack, vanishingly low. The key is making this a continuing effort, as most have said.
 
Replying to myself to add, I don't trust the cloud. I would NEVER make it my only backup.
 
Replying to myself to add, I don't trust the cloud. I would NEVER make it my only backup.
Well to each our own...:-)

Allthough I have to say that it’s such decisions that makes life harder and harder as you get older (because you start make irrational decisions like that for an increasingly number of things).

If you don’t trust the cloud, fine. Then encrypt your files before you upload them, and upload them to two different vendors - problem solved, and you are left with no work, hardware maintenance, technology shifts and so on. You can go take more pictures instead.

Just sayin´ ;-)
 
You should consider the Internet... Your life will be infinetely easier and troublefree.

Then the Cloudprovider (select a reputable one :-)) will have to handle technology shifts, dataloss prevention, multiple copies and so on... And the good ones does that on a level that FAR FAR superseedes any efforts you can do on your own. And you have the most simple interface to get your images again if you ever need them.

Just sayin’ - Life is to short to start handling hardware for your requirements.
I could not feel safe if I had only one backup in one Cloudprovider, So I should have two of them, which means double cost or have one Cloudprovider and one local backup.

I don't have much confidence in cheap and unlimited plans, so many of them suddenly stopped or changed prices.

On the other hand, for me and other people, setting up backups, NASes, networks and securing them, is just another hobby,

like you wouldn't say "life is too short to make photos of monuments or landscapes , there are thousands terrific photos osf avery site in the internet"
 
You should consider the Internet... Your life will be infinetely easier and troublefree.

Then the Cloudprovider (select a reputable one :-)) will have to handle technology shifts, dataloss prevention, multiple copies and so on... And the good ones does that on a level that FAR FAR superseedes any efforts you can do on your own. And you have the most simple interface to get your images again if you ever need them.

Just sayin’ - Life is to short to start handling hardware for your requirements.
I could not feel safe if I had only one backup in one Cloudprovider, So I should have two of them, which means double cost or have one Cloudprovider and one local backup.

I don't have much confidence in cheap and unlimited plans, so many of them suddenly stopped or changed prices.

On the other hand, for me and other people, setting up backups, NASes, networks and securing them, is just another hobby,

like you wouldn't say "life is too short to make photos of monuments or landscapes , there are thousands terrific photos osf avery site in the internet"
No, I agree. You should NEVER settle with one copy of your images. If you stop having local copies of your images you NEED to have two different cloud copies :-)
 
Optical and Magnetic media easily suffer from bit degradation, a.k.a bit rot, a.k.a. silent data corruption. Use ZFS files to guarantee your data remains exactly the same as you copied and verified. Otherwise some of your images could soon be short only showing a fraction of the total, others will have huge colorful streaks, big white blotches, sprinkles, and snow.
for my workflow, in which part should I try to implement ZFS ?

1. in my pc (dual boot win10/ Fedora Linux)
Forget about ZFS on Windows, I don't believe it is well supported. Debian based Linux distros (i.e. Ubuntu and Mint) make it easy to install the zfs-fuse package so you can read ZFS filesystems. You need a kernel extension to run ZFS in system space, which is probably the way you'd want to do it.
I mean mainly about Linux, but since it is not an of the shelf solution, it would not be easy for a novice like me
2. NAS (synology)
As far as I know, Synology and QNAP do not support RAID-Z.

Possibly a very power-efficient server could be used to make a home-grown RAID-Z server. Maybe with ARM processor, because Intel is either power hungry or slow, take your pick.
Synology does not support it, but if it was critical, I could consider building my own NAS
3. external HDDs used as off-site backups
I use NTFS for this because it is more universal than any other good filesystem (FAT32 alas does not qualify as good).
One of my 2 backups of the NAS on usb HDDs is in NTFS (the other in ext4) just in case I'll pass by the same time with the NAS, so my children could more easily restore the data, given they have the passphrase
what about other filesystems as Btrfs ?
Btrfs is supposed to be more durable than Ext4, however I think ZFS is a better option. On Linux Mint, Btrfs is a choice for Timeshift, instead of Rsync.
I'm still worried about data integrity or other failures, that may corrupt my data without me knowing that.

I forgot to mention that for important personal data (bank or tax receipts, passwords etc.) I use two different backup methods to the same medium, the first being the backup program of the NAS and the second directly from my PC to the encrypted HDD (I used Back in Time in Ubuntu until recently) in case one of these methods prove to have flaws.
Good questions.
thanks a lot for your suggestions :-)
 
Last edited:
It is really difficult to get a good estimate but reading the Wiki article on silent data loss it seems to me that in normal (life time) storage the chances are somewhere between 1 in 10^6 and 1 in 10^9 files will be lost. That means if you have two independent copies the chances of losing a file are very small indeed (say 1 in 10^12).
Most people chain their backups, so if you get a bit flip at the source or an intermediary, then it is copied down the chain to the others.

One protection here is two copies of each file on the data set. So once you've made the initial copy, you have the fall back. Zfs gives you the option to do this on a file system basis - you can have N copies of everything.

Freenas may be the easiest way to implement zfs at home. Linux zfs is something more than a hack, but still not what I'd call mature. I continue to run Solaris for this purpose.
 
If you don’t trust the cloud, fine. Then encrypt your files before you upload them, and upload them to two different vendors - problem solved, and you are left with no work, hardware maintenance, technology shifts and so on. You can go take more pictures instead.
I did not mean I was worried about file security. I am worried about files disappearing, whether by tech problems, financial problems, Russian hackers, whatever.

Actually I don't trust any single backup, no matter what kind. I figure three HDs is reliable enough.
 
"It is surprising the lack of support for the M disk storage option. "

Which makes it less attractive, there is no merit in having a backup with perfect data if it isn't easy to get the hardware to read it.
won't be a problem getting the hardware for a long time. The problem is that it's a minor PITA to deal with a multitude of discs.
 
I could not feel safe if I had only one backup in one Cloudprovider, So I should have two of them, which means double cost or have one Cloudprovider and one local backup.

I don't have much confidence in cheap and unlimited plans, so many of them suddenly stopped or changed prices.
Yes- I believe you should find the best vendor option where they charge you the proper amount for the service. Those selling below cost will eventually stop doing so, one way or the other.

But I don't think you need multiple cloud providers - you have your two versions at home. If the cloud provider goes, you use the second one at home and xfer to a new external provider. There is a pain point here getting the initial dump done, but seems like overkill to pay for 2.
 

Keyboard shortcuts

Back
Top