|
Post by christopher on Apr 30, 2019 13:24:30 GMT -6
tape is still a thing for server backup www.ibm.com/it-infrastructure/storage/tapeThanks for posting this thread, sorry to hear about this happening to you! It made me do a couple backups of stuff I forgot I needed to.
|
|
|
Post by Blackdawg on Apr 30, 2019 13:28:42 GMT -6
I use LTO7 tape for our long term storage. Shelf life of 30years.
|
|
|
Post by christopher on Apr 30, 2019 13:40:11 GMT -6
Interesting! It looks way cheaper than buying hard drives every few months. What does a starter setup cost?
|
|
|
Post by svart on Apr 30, 2019 13:42:15 GMT -6
This is terribly inaccurate. As long as your backup isn't plugged in 100% of the time and actively being written/read continuously for years during that time, it's objectively superior to spinning disks in the reliability department. Even if it was being actively used continuously, it's probably comparable reliability-wise for the first ~3 years or so. I can say that confidently as someone involved for many years with distributed systems with large volumes of storage of every type imaginable. I beg to differ. SSD drives use capacative storage, which can go bed just sitting on the shelf, unused. Granted, the odds of it soing that in the short term aree low - but a backup drive sahould be long term storage. A magnetic drive won't go bad over time sitting on a shelf, an SSD will.
And a distributed system with multiple redindancies is a very differtent thing than a backup in a small studio system. Everybody I know on the more professional forums like PRW asgree - SSDs are great for work disks because they're so damn fast, but don't use them for long term storage (backups) and always back up your work to a spinner. The only form of long term storage that rivals a magnetic hard drive is well cared for 2" tape - as long as your 2" isn't one of the types that had binder problems. Magnetic drives eventually go bad from wear, as they're electrto-mechanical systems. SSD srives go bad from "electronic rot" - there is no mechanical wear factor, which gives them an edge for short term work with many rewrites and erasures.
You're more likely to have long term problems with interface compatibility/obsolescence than with disk failure with a magnetic drive. And if the problem is with the electronics on a spinner it's a simple matter to swap in the driver boards from another drive of the same model - which you can't do with an SSD.
Also, magnetic drives have a much higher probability of being recoverable. With a magnetic drive if you have a real problem you can go in with a utility like Steve Gibson's "Spinrite" and recover stuff that your typical recovery tool (which just gathers data from sectorts marked as "unused" or "unreadable") won't even recognize as data - stuff that's several generations gone. With an SSD you can't do that because once it's gone it's gone - there are no "ghosts" hanging around haunting your platter. Spinrite accesses your disk on the rawest. lowerst level of magnetic domains - way below the level of OS readability. That does not exist with an SSD.
To be more concise, it's not quite like a capacitor in structure but it does store charge like one.. Electrons are pushed into a crystal lattice called a "floating gate" through a quantum effect called "Fowler–Nordheim tunneling". It creates a net charge difference that we measure as capacitance, but it's not like a capacitor in other aspects, like how you "charge" it. It's more like filling a bucket full of marbles by using air pressure to push the marbles through a hose. It takes significant voltage to "push" the electrons across the oxides to the floating gates where they are essentially "trapped" until another voltage is applied to erase it. unlike a normal capacitor, charge does not "leak" out at all and the only way to discharge a gate is by voltage force. Because of this, the data retention at the bit level is measured in decades, but also because of this, everytime one of these cycles happens, it physically erodes the insulation until eventually it can't retain the charge. But you also need to know that spinny drives need to be powered up occasionally for ECC to fix small errors and refresh bits as they also "lose" their magnetism slowly over time.
|
|
|
Post by Blackdawg on Apr 30, 2019 13:53:08 GMT -6
Interesting! It looks way cheaper than buying hard drives every few months. What does a starter setup cost? 15K i think. ha The machine to use them is not cheap. Holds 24 tapes but only cand read/write 2 at a time. Has little arm in it that grabs the tapes. Everything gets a bar code and meta data is stored. Its a serious archiving machine. We archive about 30TB a year so..yeah.
|
|
|
Post by christopher on Apr 30, 2019 14:06:07 GMT -6
lol.... ok thats some serious backup
|
|
|
Post by johneppstein on Apr 30, 2019 14:27:27 GMT -6
You should never use an SSD drive as a backup because they are unreliable and fail without warning. At least 90% of the time they fail they are unrecoverable - you really lucked out here.
After you recover what you can on this drive you should mark the drive bad and not use it for anything important because there is no telling when it might fail again - and the odds are you would not be so lucky. The fact that it did what it did is most likely a sign that one of the or more of the "cells" in the drive is failing and losing its ability to hold a charge over time.
You should always use a spinner for backups and archival storage.
Edit: when a mechanical drive fails more often than not it's sometyhing mechanical that prevents one or more heads from trackling reliably - the rest of the time it's a glitch on the media which can be worked around with a low level format. Since SSDs use capacitive storage that uses an array of "cells" on the chip that are determined at the factory a reformat doesn't eliminate flaws in the data structure - the bad cell will always be there, rewriting doesn't cure the problem.
SSDs have much lower failure rates than spinny drives with one caveat, lifetime. Nearer the end of their expected lifespan, they develop errors in the cells faster than HDD's do on their platters, however JUST LIKE HDD's, they have "spare" areas allocated to replace failing cell blocks. Almost - there's another exception, which is that iof an SSD DOES fail it is much less likely to be recoverable.
There is no physical wear to speak of (which is a good thing). And the chances of mechanical damage are much lower.
Which is not as useful as one might think, since most SSD failures are not gradual, like HDD failures where a sector starts getting flaky long before it fails outright. With an SSD failure is usually immediate.
Actually, in the case of a serious failure of crucial data with an HDD you shouldn't do a format at all, nor should you attempt to recover with conmventional recovery software.
I dom't know if you'rer hip to Spinrite or not (most people outside the data recovery business are not - it's very specialized program by a small, independent company that requires some technical knowledge to understand) It runs on Windows machines but doesn't use windows - it runs on a lower level than the native OS - and it's totally agnostic as to the OS used on the target disk. (It runs on its own, included version of FreeDOS.)
Usually it just means that the disk "flipped a bit" magnetically, making it unreadable to the OS. On an SSD, however, it usually means that it blew a capacitor in one or more of the storage cells on the chip, which is not recoverable in most cases.
Agreed about blanket statements about LLFs.
Not quite certain what you mean by "non-damage". Non damage to what, the disk(s)? the motor assembly? The drive electronics?
If it's the electronics you can fix that by swapping in new driver boards from another drive of the same model and rev. You can't do that with an SSD AFAIK - it's all on one big chip. If the motor is going bad and you're good enough you can install the disks into another drive and read them that way. Of course that requires having the necessary replacement hardware on hand, which most people don't - but recovery centers do.
That's a generalization. Some SSDs are more reliable than others and have much better MTBF. Aftyer 3-5 years all of them start getting unreliable. Capacitors do not have infinite lifetimes and there are a LOT of caps in an SSD.
Indeed - VERY worth noting. Not all SSDs are created equal and it's still a developiong technology, which is an excellent reson for not relying on it for things like storage of important data. Great work drives, yes. Great system drives, sure - blazing fast and you should always be able to replace your system drive from the installation files. But not good for backups and long term storage.
Your dragster may do a blazing quarter mile but it's not a very reliable vehicle for commuting to work and doing shopping.
Tape still is (for audio), provided it's stored properly and you keep a well maintained machine tp play it. And your tape isn't a bum formulation likle 456.
|
|
|
Post by m03 on Apr 30, 2019 14:37:53 GMT -6
I beg to differ. SSD drives use capacative storage, which can go bed just sitting on the shelf, unused. Granted, the odds of it soing that in the short term aree low - but a backup drive sahould be long term storage. A magnetic drive won't go bad over time sitting on a shelf, an SSD will. Long term storage is relative, and IMHO you should be migrating to something newer periodically (every decade-ish), from the perspective of logistics. As you alluded to, it requires a non-trivial level of effort to get those backup files off of you 35 year old 8-inch spinning disk drive that relies on an external controller that was obsoleted 30 years ago, from a proprietary vendor that went out of business 20 years ago...assuming it's still usable at all since you likely didn't store it in a climate controlled environment, and you definitely didn't power it up periodically. This is a problem universal to storage mediums. And a distributed system with multiple redindancies is a very differtent thing than a backup in a small studio system. Missing the point. I've had a close eye on the reliability differences since the inception of SSDs, and in practical applications, spinning disks die much more frequently, and they're much more sensitive to environmental issues (oops, I dropped my backup drive, oops, I accentually put my backup drive on top of the magnet that I forgot was in that drawer, oops, I left my backup drive in the car the whole summer, etc). Your other points are more or less irrelevant from the perspective of a drive that is dedicated to being a purely backup solution, as SSD will be more reliable even if only considering environmental factors alone. Of course, and this point in the evolution of technology, Amazon Glacier and similar storage services are probably the wiser move.
|
|
|
Post by svart on Apr 30, 2019 15:11:19 GMT -6
SSDs have much lower failure rates than spinny drives with one caveat, lifetime. Nearer the end of their expected lifespan, they develop errors in the cells faster than HDD's do on their platters, however JUST LIKE HDD's, they have "spare" areas allocated to replace failing cell blocks. Almost - there's another exception, which is that iof an SSD DOES fail it is much less likely to be recoverable.
There is no physical wear to speak of (which is a good thing). And the chances of mechanical damage are much lower.
Which is not as useful as one might think, since most SSD failures are not gradual, like HDD failures where a sector starts getting flaky long before it fails outright. With an SSD failure is usually immediate.
Actually, in the case of a serious failure of crucial data with an HDD you shouldn't do a format at all, nor should you attempt to recover with conmventional recovery software.
I dom't know if you'rer hip to Spinrite or not (most people outside the data recovery business are not - it's very specialized program by a small, independent company that requires some technical knowledge to understand) It runs on Windows machines but doesn't use windows - it runs on a lower level than the native OS - and it's totally agnostic as to the OS used on the target disk. (It runs on its own, included version of FreeDOS.)
Usually it just means that the disk "flipped a bit" magnetically, making it unreadable to the OS. On an SSD, however, it usually means that it blew a capacitor in one or more of the storage cells on the chip, which is not recoverable in most cases.
Agreed about blanket statements about LLFs.
Not quite certain what you mean by "non-damage". Non damage to what, the disk(s)? the motor assembly? The drive electronics?
If it's the electronics you can fix that by swapping in new driver boards from another drive of the same model and rev. You can't do that with an SSD AFAIK - it's all on one big chip. If the motor is going bad and you're good enough you can install the disks into another drive and read them that way. Of course that requires having the necessary replacement hardware on hand, which most people don't - but recovery centers do.
That's a generalization. Some SSDs are more reliable than others and have much better MTBF. Aftyer 3-5 years all of them start getting unreliable. Capacitors do not have infinite lifetimes and there are a LOT of caps in an SSD.
Indeed - VERY worth noting. Not all SSDs are created equal and it's still a developiong technology, which is an excellent reson for not relying on it for things like storage of important data. Great work drives, yes. Great system drives, sure - blazing fast and you should always be able to replace your system drive from the installation files. But not good for backups and long term storage.
Your dragster may do a blazing quarter mile but it's not a very reliable vehicle for commuting to work and doing shopping.
Tape still is (for audio), provided it's stored properly and you keep a well maintained machine tp play it. And your tape isn't a bum formulation likle 456.
I used a bit-level recovery tool some time ago when a spinny drive I had went bad with some crucial data on it. The drive itself lost the converter chip between the headstack and the driver board so the drive itself would have been completely lost, but I did take a huge chance and bought another identical drive with the same firmware revision and proceeded to swap the platter in it. It was a risky move, but in the end the new drive "worked" but windows would not see it as initialized. I then used the program to "find" the data, which it did remarkably well and I don't think I actually lost any data at all. I can't remember the name of the tool but it was donation-ware from some guy who wrote it to recover his own files. DOS-based of course. Saved my ass for sure when during all this I found out my company's backup system hadn't actually backed my machine up for some time due to someone not entering in my IP correctly.
|
|
|
Post by johneppstein on Apr 30, 2019 21:16:24 GMT -6
I used a bit-level recovery tool some time ago when a spinny drive I had went bad with some crucial data on it. The drive itself lost the converter chip between the headstack and the driver board so the drive itself would have been completely lost, but I did take a huge chance and bought another identical drive with the same firmware revision and proceeded to swap the platter in it. It was a risky move, but in the end the new drive "worked" but windows would not see it as initialized. I then used the program to "find" the data, which it did remarkably well and I don't think I actually lost any data at all. I can't remember the name of the tool but it was donation-ware from some guy who wrote it to recover his own files. DOS-based of course. Saved my ass for sure when during all this I found out my company's backup system hadn't actually backed my machine up for some time due to someone not entering in my IP correctly. Understand, Spinrite is neither "bit level" nor is it freeware/donationware (it costs about 90 bucks). It has been a "secret weapon" of profesional data recovery people since the '90s (f not before), going through several levels of revision.
It functions on the level of the actual magnetic domains on the disk.
It's been a long time since I used Spinrite (and I only did so a couple of times), as I try to avoid doing computer tech these days (unlike the period from 1980 through about 2007, give or take a couple of years.) However if you do much that involves data recovery I'd strongly recommend checking it out.
|
|