Internal SATA hard-drives reliable?

Hello everyone,

I'm sitting on about 2-4 TB of data that has never really been backed up properly. Now, by properly I mean something reliable. Reliable is where I know that the data won't be corrupted for the next 20-30 years or so until I can back it up into another better medium. My question is, are SATA hard-drives that reliable? What about tape drives?

Side note: after I decide on the medium, I might be going with the ZFS route.
 
mrkotfw said:
Reliable is where I know that the data won't be corrupted for the next 20-30 years or so until I can back it up into another better medium.
Do you really mean 20 years? In 1992 a 1GB drive in a 5.25" form factor with a synchronous SCSI interface was cutting edge. In 1982 a 300MB drive in a 19" rack with a SMD interface was cutting edge.

I expect that there are still some Fujitsu Eagle drives still operating. Those drives were indestructible - 10 years after I'd installed one I was relocating the rack it was in and noticed (for the first time) that it had an air filter that was supposed to be cleaned regularly.

Anyway, the point is that your 3.5" SATA drive is likely to be completely obsolete well before 20 years. You can already get portable USB 2TB drives - why not just make a copy of your data this decade?

My question is, are SATA hard-drives that reliable? What about tape drives?
3.5" hard drive reliability over a period of decades is unknown. You'll have 2 types of failures to worry about - a drive in continuous operation failing (if 5.25" drives are any indication, probably from a spindle bearing failure) or a drive that's been powered off in storage failing to go ready when powered on for the first time in years. Out of a population of over two hundred 3.5" hard drives that were more than a decade old (SCSI or PATA - SATA didn't exist then) that I tried to write 0's to, only a few (less than 5%) were unable to write the whole drive. Some of the ones that didn't work may have been bad when they were put into storage more than 5 years ago. Some of the bad ones may have been recoverable by replacing the boards (allowing recovery of the data), but since my goal was to wipe them, I didn't test this.

I've read dozens of 15-year-old DAT tapes (recently). I can also read other tape formats going back to 9-track 800BPI, including some written over 30 years ago. But a good number of the older tapes have become unreadable, usually due to the oxide coming off the backing or the backing becoming sticky. The biggest problem with tape drives is that if you have one just sitting there powered on, the fan will be continually depositing dust in the mechanism. This can cause damage to tapes (and sometimes the drive) when you try to read them.
 
  • Thanks
Reactions: lbc
What about writing Blu-rays? Once they get cheaper and with more capacity (I haven't seen 100GB yet), they should be more resilient than both disks and tapes.
 
Twenty years ago my backups were on 2GB DAT and DEC TK50 tapes. I don't have any of this media anymore, nor do I have drives that would read them. I still have some stuff from that era that made it to CD-R, DVD+R and current hard disks. A few things other people have archived on the web (e.g. DECUS stuff).

bbzz said:
What about writing Blu-rays? Once they get cheaper and with more capacity (I haven't seen 100GB yet), they should be more resilient than both disks and tapes.

Current media are 50GB, so 4TB would be on the order of 80 discs or more. The going price I found on the Delkin "200 year" discs is about $10 each, so the media will be $800 or more. And I sure wouldn't make a multi-disc backup set that would require reading an average of 40 discs to find the files you want, either, so you will have to do some work to split up and organize everything into individual discs or small backup sets. Even 100GB media will give problems of scale, but only half as bad.

The question of whether the discs will really be readable in 20 years for a reasonable cost is still open, despite manufacturers' claims.
 
It will cost less and be more reliable and accessible to plan to update backup strategies every five years. Otherwise it's more of a data recovery operation than a restore. Carefully store a BD writer from today and it still may be dead in twenty years: drive belts deform, grease in the bearings, electrolytic capacitors, all can go bad in storage. If it still works, will it be able to connect to the 1.3V UltraMegaIO ports that have become the standard? Once all that is settled, is there any software that still understands the file format on those disks?
 
  • Thanks
Reactions: lbc
Terry_Kennedy said:
Do you really mean 20 years? In 1992 a 1GB drive in a 5.25" form factor with a synchronous SCSI interface was cutting edge. In 1982 a 300MB drive in a 19" rack with a SMD interface was cutting edge.

I expect that there are still some Fujitsu Eagle drives still operating. Those drives were indestructible - 10 years after I'd installed one I was relocating the rack it was in and noticed (for the first time) that it had an air filter that was supposed to be cleaned regularly.

Terry_Kennedy said:
Anyway, the point is that your 3.5" SATA drive is likely to be completely obsolete well before 20 years. You can already get portable USB 2TB drives - why not just make a copy of your data this decade?

That's something that I hadn't realized. At the verge of the 3.5" SATA drives becoming extinct, I should still be able to read from the drives without any corruption. The idea that I have now is that I should be able to move my data from one medium to another without actually losing any of that data.

Terry_Kennedy said:
3.5" hard drive reliability over a period of decades is unknown. You'll have 2 types of failures to worry about - a drive in continuous operation failing (if 5.25" drives are any indication, probably from a spindle bearing failure) or a drive that's been powered off in storage failing to go ready when powered on for the first time in years. Out of a population of over two hundred 3.5" hard drives that were more than a decade old (SCSI or PATA - SATA didn't exist then) that I tried to write 0's to, only a few (less than 5%) were unable to write the whole drive. Some of the ones that didn't work may have been bad when they were put into storage more than 5 years ago. Some of the bad ones may have been recoverable by replacing the boards (allowing recovery of the data), but since my goal was to wipe them, I didn't test this.

Thanks, I didn't think having drives powered off for a long time would also damage them.

Terry_Kennedy said:
I've read dozens of 15-year-old DAT tapes (recently). I can also read other tape formats going back to 9-track 800BPI, including some written over 30 years ago. But a good number of the older tapes have become unreadable, usually due to the oxide coming off the backing or the backing becoming sticky. The biggest problem with tape drives is that if you have one just sitting there powered on, the fan will be continually depositing dust in the mechanism. This can cause damage to tapes (and sometimes the drive) when you try to read them.
 
What was always recommended to me was to copy the backup onto another, newer, medium every 5 years or so. This ensures that you will at least always be able to read the latest copy. This still does not ensure that you can actually get to the data that is within the files or anything like that.

20-30 years is a very long time, very little information that I have will be of any use in the range of time. If I needed things to be available then I would probably print them out and store a paper copy of them somewhere with controlled humidity and no sunlight.
 
Make two backups

I agree the only way you can insure data is to print it out on paper and store it correctly, that has worked for over 2000 years.

I had to retrieve a file from a 1991 1GB Micropolis SCSI drive early this year, that drive had been on a shelf in a closet for about 15 years,. I bought a used adaptec disk controller and popped it into the PCI slot of my computer and loaded the cards driver. I powered the system down, hooked up the drive which was sitting on a clipboard laid across the case which was sitting on it's side. I snapped on the power and all that data was there and the drive sounded fine, I ran Spinrite on the disk and it reported the drive was in good shape. That was atop rated drive in it's day, I doubt you buy that kind of reliability these days. Tape is a bad road littered with potholes, I've seen too much oxide shed off too much tape to trust it.

Twenty-two years ago I paid $980 for that drive, I'd be surprised if the $109 WD 1.5TB drive I just bought is working in 5 years. If you want to be safe buy two sets of enterprise class drives and write the data to them and store them someplace that has stable temperature and humidity. Then you have to guard against the fact that the hardware standard may not be around in 5-10-15 years so you will have to do it all over again with the then current standard "disk" hardware.

Printing out all that data saves you from that worry but typing it all back in would be quite a job!

Bob
 
bobc4775 said:
...the only way you can insure data is to print it out on paper and store it correctly, that has worked for over 2000 years.

Not exactly. As far as I know the oldest surviving book is about 2500 years old, but it is written in Etruscan characters and the pages are made of 24 carat gold. And it might not be the most important/significant work of the period, but it is the only survivor that has not been recopied (and possibly changed) many times. And BTW we can't read it because we don't speak the language. The Derveni papyrus is about 2350 years old, but is not in very good condition. I think the story of the Archimedes Palimpsest is pretty interesting, and it illustrates the state of affairs in truly old books.

I doubt much that is printed on the typical modern cheap copy/printer paper will be readable in 100 years, let alone 1000. Certainly paper made with cotton bond fibers, or other real rag content, will last much longer than paper made from wood pulp. Most really old books are ink on vellum (made from sheep or other animal skin).

Another major problem with the idea of typing data back in from printed copies are the huge number of errors that are likely to result. I really don't see this being a workable approach for arbitrary data. It barely works for beautifully written prose or poetry that the copyist really believes in.

Hopefully someone will eventually develop a true archival quality digital medium with sufficient information density that it will make sense to save stuff that way.
 
Yeah, barcodes work at some level. So do two-dimensional codes (QR or some derivative). But how much paper will a terabyte take? I just don't see this being the answer for arbitrary data.
 
The point about the quality of paper used is valid, cheap copier paper would be lucky to last a century if stored properly. Barcodes are interesting but reading them in a dozen centuries might not be easy. I've had enough failures of CD-R's and DVD-R's to know that is not the way to go, I've never tried the archival types but I wonder if they are any better than a hard disk, you can buy a good quality 2 TB disk for a lot less than the number of good quality recordable DVD's. With either one you would have to copy them over to new media every five to ten years.

I read an article a few years ago citing the problem of long term data storage and they cited the problem of having a working device to read the data back, the consensus was good quality paper was the only sure bet for truly long term storage IF it was done properly. We could take a page from NASA and use a real gold stamping of a record or laser disk, that would be very expensive but it would not have to be copied for a very long time. In the case of a laser disk the problem of playback in a few centuries would be a problem.

Bob
 
Magnetic storage is on its last legs. There's a great demand for something truly new in the area of storage or the SSD technology needs to take some huge steps forward to be able to replace magnetic storage.
 
At the end of the day, whatever media you are using (disk, tape, SSD, whatever) - bet on it dying and take precautions to mitigate the consequences. That typically means multiple copies of the same data, whether that is live copies via some form of RAID (to maintain availability) and several archive tapes, etc.
 
A few random observations now that I've re-read the whole thread...

I have a lot of old files still available on spinning disks. The oldest file date that I trust is from February 1980, with a system lineage as follows: CP/M system -> IBM PC clone -> VAX/VMS -> Alpha/VMS -> BSD/OS -> FreeBSD. The oldest file I'm serving via anonymous FTP / HTTP dates from early 1989, if not earlier (there are terabytes worth of that data, and I'm not going to dig through it to find an earlier date). That was probably a timestamp of when it was copied, not created, since a number of these passed through FTP along the way.

I have printouts of my work dating back to 1974. These early ones were from an ASR33 Teletype, which used very low-quality paper. Aside from some yellowing, they're still fine. Slightly newer printouts are on various grades of pin-feed paper, printed by DECwriter II, Trendata 1000, and similar terminals, as well as some IBM mainframe line printers. They all seem legible and not fragile, other than being somewhat brittle at the fan-folds between pages.

I have paper tape and punched cards from the same era and have no reason to believe they wouldn't be readable if I had the hardware to read them. I have some punched Mylar tapes which were intended to last "forever" and I have no reason to believe they won't make it to 100.
 
Finding hardware to read old media can be hard. I noticed the drive I pulled out of my current-2 generation machine (a Pentium-D 930 that I just decomissioned) was running PATA hard drives. New ports only have SATA ports. Floppy drives are hard to come by. DVDs are not long for this world.

My point is that even if the media survives, a machine capable of reading it may be hard to find. Keeping archive media stored aware is probably not enough. You need to regularly re-write to new media as formats are deprecated.
 
Back
Top