Magnetic decay in modern hard drives.

[wanneslier]

When you back up data and leave them (on a hard drive) for a considerable amount of time, chances are that some bytes may get coorupted.
Therefore it is a good idea to rewrite a hared disk with stable data on it every now and then.
So I am looking for a software that rewrites a certain percentage of a drive every day, so that it is rewritten e.g. every 200 days (rewriting 0,5% of the drive every day).

Does anyone know of such a program?

The more one leaves on media that will get corrupted but are not read-once, the more interesting this procedure gets.
So as more and more data are "stable" on a hard drive (or in the future on an SSD), it becomes really important to make sure the data is preserved in its original state ...

[RedSon]

Is there a particular problem that you are trying to solve?

Drive manufacturers design their drives to prevent this degradation by flagging corrupted sectors. You can prevent data loss by using a journaled file system and defragging your machine every so often.

If you have particularly important information you might want to switch to a RAID system.

[wanneslier]

Not really a solution. All those are correct on an "active" drive. I'm talking about a drive where e.g. music files are put. Once the drive is full, it is left in the system to play music from, but never rewritten. After some time (years, I know, but that doesn't cure the problem) written bytes, because of loss of magnetic intensity, are "lost" in that they cannot be read correcdtly anymore. This has nothing to do however with sectors being corrupted or physically faulty. To prevent this loss, one can read a small part of the drive every day and rewrite it, thereby checking that the read and written parts are identical. This way one is sure the slow degradation of magnetic intensity is overcome.

So it is not about faulty or corrupted hard drives, it is about the basic workings of hard drives.
In fact the same holds true for CD's or DVD's (especially of lesser quality) who loose their information integrity after maybe 10-15 years depending on how they are kept.

That is why I'l looking out for a software that does exactly that: rewrite small portions of a disk on a daily basis so that one is sure the information will not be lost because of magnetic intensity loss.

[arnaudk]

Modern hard drives bigger than ~1GB don't store information in classical magnetic domains like cassette tapes. Instead, the information is stored by aligning quantum spins through a phenomenon known as giant magnetoresistan ce (in fact GMR was the subject of last year's Nobel prize in physics). The decay time for adjacent spin domains probably in the order of decades at least in modern materials, much longer than the decay of classically magnetized domains (what you refer to as "magnetic intensity loss"). Thus "reinvigorating " your data is hardly necessary with modern hard drives which is probably why nobody has written such a piece of software.

Now, optical storage media is an entirely different matter of course, as is the corruption of data through manufacturing tolerances, external interference, material imperfections etc. Reinvigorating your data through software won't save you from those.

[RedSon]

Modern hard drives bigger than ~1GB don't store information in classical magnetic domains like cassette tapes. Instead, the information is stored by aligning quantum spins through a phenomenon known as giant magnetoresistan ce (in fact GMR was the subject of last year's Nobel prize in physics). The decay time for adjacent spin domains probably in the order of decades at least in modern materials, much longer than the decay of classically magnetized domains (what you refer to as "magnetic intensity loss"). Thus "reinvigorating " your data is hardly necessary with modern hard drives which is probably why nobody has written such a piece of software.

Now, optical storage media is an entirely different matter of course, as is the corruption of data through manufacturing tolerances, external interference, material imperfections etc. Reinvigorating your data through software won't save you from those.

Hi,

Can you please site some of your sources. It's not that I don't believe you it is that I would like to read more about it.

Thanks.

[arnaudk]

Hi,

Can you please site some of your sources. It's not that I don't believe you it is that I would like to read more about it.

Thanks.

I didn't use a specific source, spintronics is close to one topic I have researched (the 3D Heisenberg model in the context of ultracold atoms) but I think wikipedia has a lot of information on it. My knowledge of the GMR comes comes mainly from the following notes: http://www.phys.uu.nl/~duine/Rembert...pintronics.pdf Let me know it there's anything specific which you'd like to know more about and I'll do my best to find some open access references on the topic.

[arnaudk]

Here are some popular articles I found on GMR:

http://www.popularmech anics.com/technology/industry/4226779.html

http://www.research.ibm.com/research/gmr.html

Wikipedia

[wanneslier]

The decay time for adjacent spin domains probably in the order of decades at least in modern materials, much longer than the decay of classically magnetized domains (what you refer to as "magnetic intensity loss"). Thus "reinvigorating " your data is hardly necessary with modern hard drives which is probably why nobody has written such a piece of software..

Thanks a lot for the information. Still ... "probably in the order of decades" ... how probable is this and how many decades. The reason I'm so persistent is simple: since I began using PC's (start of the 80's) I've lost so many material, learning the hard way to better use backups, that I want to minimize my risk. So I already use mirrored drives (fully mirrored) to avoid as much as possible any loss of data due to HD crashes. In your answer the wqrd "probablty" is what makes me shiver. It's the same word that was used with backups to CD's quite some years ago - and which got me into trouble on specific data.

So can you elaborate on "probably in the order of decades"? And 10 years are gone quite fast - which brings us again to the "reinvigorating " of my data ...

I very much appreciate the feedback on this forum. Thanks guys !!

[RedSon]

If you are looking for serious time spans then you should get these Verbatim UltraLife Gold Archival Grade DVD-R

I heard they are supposed to be good for 100 years.

[Curtis Rutland]

I honestly don't see the point of this. Your HDD is far more likely to fail for other reasons long before this ever comes into effect.

But that's the reality of computers...HDD s fail. They're not the best archive solution out there, but for the average consumer, the combination of an HDD and burning DVDs is more than enough. If you are really that worried about your data, buy a tape drive, and lock the tape in a safe.

Also, you have to consider that you will be buying newer HDDs/SSDs before too long. To put it another way, your disks may outlive their usefulness to you. Because eventually, a new disk will be cheap enough with enough storage space and benefits that you will buy it and move your data to the new one. Plus, interfaces change. IDE HDDs are already going the way of the dodo. Eventually SATA will too. So, there are many, many other things to consider before you worry about static data being lost from an HDD.

Perhaps the reason that you aren't finding the software you want readily available because there is no need for it. If there was, you would be sure someone would be selling it.

[RedSon]

Best archival media is stone tablet...

[Curtis Rutland]

Yeah, but the read/writes are pretty slow...

[arnaudk]

Best archival media is stone tablet...

Cave wall painting is pretty reliable too. Some clever Cro-magnons archived their holiday snapshots on the walls of the Lascaux cave 16,000 years ago and they're still intact. But it's hard to find a good cave nowadays...

[wanneslier]

:-))

Ok - sorry for the drag
By the way I bought DVD-R's from Tayo Yuden who are "guaranteed " for a 100 year lifespam.

So I'll look out for a new external drive before long ...
whatever the interface, 1TB is going below €100 these days, so what the heck.

Thanks for the help and I'll have a close look at the caves. Alternatively, as my wife is a painter and is experimenting making her own colors, I'll see to it they last a cuople of thousand years, but then the canvas should be good enough (papyrus is quite ok I suppose)

[arnaudk]

I did a little searching on the details how how HDDs demagnetize. From what I can ascertain, it looks like it would take about 22 years (details below) for you to loose your data due to thermally-driven demagnetization if the hard drive were just sitting motionless at room temperature in a dark corner. In reality, this time will be a bit shorter because of mechanical vibrations and external magnetic fields arising due to everything from the motor of the hard drive itself to lightening storms 50km away. Still, (aside from the fact that the data is usually regularly rewritten anyway) in the end, the hard drive will fail long before it demagnetizes due to other wear-and-tear factors so you don't need to bother re-invigorating your data (technically this is measured as the Mean Time to Data Loss [ref0]).

---Details:---
The magnetic domains that store bits on modern hard disk platters are comprised of a large number of magnetic grains. Although the entire bit domain can have a varying degree of magnetization strength, a grain is either magnetized or not magentized. The time it takes for a grain's magnetization to flip or decay is determined by the Néel-Arrhenius equation.

Now, for Cobalt-Chromium alloy magnetic films typically used in modern hard drives, the energy barrier is Eb = 2 x 10^5 J/m^3 * V and grain diameter is 10nm diameter yielding a volume of roughly V= 5.24 x 10^-25 m^3 [ref1]. Thus, Eb/(kB*T) ~ 25 where kB=1.38x10^-23J/K is Boltzman's constant and room temperature is T = 300 Kelvin. With a typical attempt frequency of 1GHz [ref2][ref3], this leads to a decay rate of 10^9 * exp(-25) = 0.01 grains/s = 326000 grains/year.

Acceptable levels of signal decay vary depending on system design but typically range between 10-20% [ref4], so it would take (-1/326000)*ln(0.8) = about 22 years for an entire bit domain to get 20% weaker causing possible loss of data due solely to thermal demagnetization effects.