Home
Posted By: jchuzi SSD vulnerability? - 05/18/15 10:37 AM
Any thoughts about Leaving unpowered SSDs in a warm room can kill your data fast?
Posted By: Ira L Re: SSD vulnerability? - 05/18/15 02:06 PM
I saw that article and the title seemed to be scarier than the actual facts: small data sample, extreme parameters, etc. Still, it does raise an interesting question regarding quantum physics that may become a significant issue down the road.
Posted By: tacit Re: SSD vulnerability? - 05/19/15 03:08 AM
There is discussion of this report over at Slashdot, and some of the people commenting there have actually seen SSDs losing information when left alone for a while.

I have an iMac desktop system that I keep in my writing office, which is a loft over the house. I'm on the road a lot and bring my laptop, so my desktop system will often go as long as two months without being powered on. The loft isn't air conditioned when I'm not home. In that kind of scenario, an SSD, from the sound of it, might be dead pretty quickly.

SSDs (well, NAND flash memory generically) are terrible at retaining data. Imagine each bit as a slightly leaky balloon; the disk controller normally "walks" the drive slowly when it's idle, refreshing the cells (note: this is totally different than the kind of refresh that is done in dynamic RAM) to prevent information loss. But it's totally possible for a consumer to have a computer that's off for a long time, then turned on to get something--say, an old machine that has a tax return from two years ago on it. Surprise! It's actually quite likely that a system left turned off for months or years will not be able to read the information that used to be on its SSD.
Posted By: jchuzi Re: SSD vulnerability? - 05/19/15 09:02 AM
Since my iMac has a fusion drive, I wonder how this issue will affect it. My computer is turned off only at night, so I'm not too worried about it.
Posted By: Virtual1 Re: SSD vulnerability? - 05/19/15 11:45 AM
maybe they should be re-investing some of that space and r/w performance into error correction?
Posted By: tacit Re: SSD vulnerability? - 05/19/15 08:30 PM
They already do. SSD devices have a surprisingly large amount of "noise"--a lot of the memory cells don't return a definitive 0 or 1 when read, but an indeterminate state that must be reconstructed by error correction hardware and software into a 0 or 1.

The fundamental problem is that the memory cells don't hold their state. For example, NAND flash memory, the most common type of flash memory, is a "floating gate" device. There's a sliver of conductive material suspended in an insulating layer that's narrower than the wavelength of an electron. You can get a charge into it, even though it's surrounded by an insulator, by exploiting a weird quantum mechanics effect called "electron tunneling," whereby a particle can travel from one place to another without passing through the space between.

Unfortunately, electrons in this gate can tunnel right back out. It's a probabilistic process, and over time, more and more of them will tunnel out, until eventually the bit at that address changes. There's no way to prevent it; it's a fundamental property of quantum mechanics. Error correction circuitry can help to a certain extent, but once enough bits are lost, there's no way to recover the data.
Posted By: joemikeb Re: SSD vulnerability? - 05/20/15 12:26 AM
This strikes me as some reporter looking for a sensational issue to garner attention and headlines. SSD and similar devices will lose memory over time and that memory loss may be accelerated by heat. That is equally true of magnetic storage media and devices such as hard hard drives — but how concerned are we about their data loss?

If this were a really serious issue then not only SSDs but also thumb drives, SD cards, iPods, iPhones, iPads, MacBook Airs, Apple Watches as well as smart phones, tablets, etc from other makers would be relatively useless if not totally unusable. All use SSDs of one form or another and by their very nature are often subject to heat and/or go unused for long periods of time.

Across the entire southern and southwestern areas of the United States the interior of a parked vehicle on a summer day can easily reach 140 to 150 degrees and the entertainment systems, GPS units in modern cars all have NAND memory devices not unlike those in SSDs and memory loss does not seem to have been an issue with them.
Posted By: Virtual1 Re: SSD vulnerability? - 05/20/15 11:49 AM
the way to correct for "bit-rot" is to conduct "bit-refresh". Where you read the media, and when you find a single bit error that you can recover via EC, you change the bit back to the state it was supposed to be.

(of course this only works if the device is powered on and has an opportunity to "creep" over the media in its idle time)
Posted By: tacit Re: SSD vulnerability? - 05/23/15 01:35 AM
I think it is a legitimate issue. Yes, the same issue also applies to "spinning rust" hard drives, but much more slowly--I've had no problems reading old-fashioned hard drives that have been sitting in the loft for 7 years. An SSD would be very unlikely to retain its information that long.

It's also true of iPads, iPhones, and thumb drives, but people don't usually leave those turned off for long periods of time, and in any event aren't likely to keep critical data on them.
Posted By: joemikeb Re: SSD vulnerability? - 05/23/15 04:06 PM
Originally Posted By: tacit
It's also true of iPads, iPhones, and thumb drives, but people don't usually leave those turned off for long periods of time, and in any event aren't likely to keep critical data on them.

Agreed. I keep most of the critical data used on my iPhone and iPad safely tucked away on iCloud. grin
© FineTunedMac