Home Forums Peripherals & Multimedia Peripherals "Exercising" HDs and SSDs used for storing data

YIKES!

Back in the mainframe days it was standard operating procedure to "exercise" — read and rewrite — magnetic tapes on a regular basis to prevent data deterioration and/or loss. Since everything has pretty much migrated to HDDs and more recently SSDs (which includes flash drives, SD cards, and other similar solid state devices) that had pretty well gone out of my consciousness until I came across this Rocket Yard article from OWC. Enterprise class SSDs with a data retention rate of only three (3) months and Consumer grade SSD retention rates of only one (1) year.

I have a lot of data stored on solid state devices of one sort or another, including a Drive Genius Bootwell device and a TechTool ProToGo device kept for emergency repairs. I am going to have to institute a regular program of "exercising these devices or forego the SSD speed and compact size advantages and revert to HDs when devices are seldom used.

Interesting article. I would note that the SSD degradation times mentioned by you and in the article are for devices stored at 30-40 degrees C., which tends to be warmer than most domestic or commercial interiors.

The "scary" observation mentioned for revolving hard drives is that it is not enough just to power them on. It is recommended that full re-writes via copying or cloning back and forth should be done. YIKES is right.

I agree the temperatures cited are higher than one would expect in a home or business environment and as the article stated lower temperatures retard deterioration. Back in the magnetic tape days the company I was with trucked their tapes 400 miles to a limestone cave in Arkansas for storage because the cave was cool and dry, but even those were "exercised" biennially. (Locally stored tapes were exercised at least twice a year.)

Even if the SSD tests were run under less than ideal conditions (86 & 104℉) we are still looking at data reliability in terms of months or a very few years. Hard drives may be slow, but at least their data retention is measured in decades — admittedly under better conditions. Unless there is a Solid Sate breakthrough our Hard Drives will be with us for a long time if for no other reason than the data on them will be there for a longer time than on the SSDs.

I hadn't thought of this before, but this does make another good case for using online data storage such as iCloud.

Does anyone have a suggestion for reliable and reasonably priced online archives?

Re: " I hadn't thought of this before, but this does make another good case for using online data storage such as iCloud."

---------------


Assuming, in the first place, one is comfortable with storing business or personal data in iCloud. I certainly wouldn't (willingly) store anything of a sensitive - or can't afford to lose - nature.

even though it's a bit alarming, this isn't really very surprising. Flash memory stores information as a static charge, and this charge will degrade over time. It's not a physical switch being flipped. It's more like a tiny battery being charged up to indicate a "1", or NOT charged up to indicate a 0. Billions of little batteries.

So over time they slowly discharge, and the 1's slowly creep toward being 0's. Given enough time, some of them actually start to cross that "line in the sand" the hardware has set for calling something a zero. Given enough time, all the 1's change back to 0.

If you read the bit, and then write it, that will restore the full charge to any 1's you read, and will "reset" them back to their maximum charge / lifespan.

I would expect any higher quality flash drive to periodically 'sweep' itself and refresh the charge on 's but I don't know if that actually happens. And I don't know of any software that does it. Something like this also looks like something that the flash controller chip may actually interfere with. Part of its job is to prevent unnecessary writes, because they are (almost?) always preceded with a wipe of the block (back to all zeros) before writing. So I could see them taking any of a number of approaches to writing data to a block that's already got data in it.

A- read it (takes time) and see that it's the same (no change) and decide "nope! we don't need to touch this" and just ignore the write completely
B- read it (takes time) and see that it's the same (no change) and decide "to skip the wipe and just do a re-write refresh
C- just immediately wipe+write it in place
D- remap the block to another cell and wipe+write it (to reduce wear on the original cell)

I would expect most controllers would opt for D, although what you appear to be looking for is actually B. This would wear out the drive faster. But the reason I think they would do this is because they're just not anticipating people intentionally refreshing their disk images.

Alarming: No. A precautionary heads up: Yes.

If there is a flash drive with an automatic “refresh” you would think it would be a major marketing point but I have not seen or heard of it and apparently neither has OWC.

To my personal knowledge, this issue has been researched/discussed/argued since 1950 and many white papers, and scholarly articles written about it. The DoD spent tens of millions of dollars trying to find acceptable technologies (remember “Bubble Memory”?) and in the end there is still no cost efffective solution other than periodic “refreshing”.

Your idea of built in automatic refresh is not new and was used in Magnetic Core RAM and is still used in Solid State RAM but unlike flash media RAM does not have a finite write life. Our salvation today is the relatively short lifespan of modern computers. The computer is more likely to become obsolete and discarded before the SSD exceeds its write limit. Of course that still leaves us in need of a good archival storage media and there aren’t any in site.

It doesn't surprise me that anything that's fast and easy to write to at high density will inevitably be a poor choice for archival storage. The design requirements are polar opposites.