Slashdot Mirror
No, Your SSD Won't Quickly Lose Data While Powered Down
An anonymous reader writes: A few weeks ago, we discussed reports that enterprise SSDs would lose data in a surprisingly short amount of time if left powered off. The reports were based on a presentation from Alvin Cox, a Seagate engineer, about enterprise storage practices. PCWorld spoke to him and another engineer for Seagate, and they say the whole thing was blown out of proportion. Alvin Cox said, "I wouldn't worry about (losing data). This all pertains to end of life. As a consumer, an SSD product or even a flash product is never going to get to the point where it's temperature-dependent on retaining the data." The intent of the original presentation was to set expectations for a worst case scenario — a data center writing huge amounts of data to old SSDs and then storing them long-term at unusual temperatures. It's not a very realistic situation for businesses with responsible IT departments, and almost impossible for personal drives.
yeah, right.
Backup your stuff. Make it a part of your ritual. Be the data.
I'm interested to note this source was a representative of Seagate.
That makes sense, every Seagate drive I've ever owned lost a huge amount of data whenever it was powered off. Or left on.
based on a presentation from Alvin Cox, a Seagate engineer[...]Alan Cox said, "I wouldn't worry"
Can we get these two gentlemen to agree on a statement of risk? Or maybe just a little, you know, editing from the Slashdot editors?
I have tried SSD's since they were a new thing. I'm a software developer. To this day I have never seen one last more than a year under my "abusive" conditions. Yes, they are awesomely fast but longevity is just not their thing.
I still have a normal old spinning hard-drive keeping my most important data. This drive has over 83000 powered-on hours (according to SMART). In the time I have had this drive I have replaced half a dozen SSD's that weren't really used for much other than OS data. What does that tell you?
We have a long-term (over several years) exercise where we collect large amounts of data at a very hot waterside location overseas, in an air-conditioned office. And then the racks get moved to a non air-conditioned warehouse where they sit for a year or longer. And then we come back after a year or longer, move the racks back into office space, and do the next iteration.
We were thinking about going to SSD, just for the drive performance, and we now know that our setup is a poster child for the problem that the alarmist article described.
Of course, we copy off the valuable data and take it home, but coming back to random corruption on our system and middleware drives could introduce some real issues.
I have several personal drives in a storage shed, that end up there as secondary backups when I upgrade the drive in a system. It's always about 5 degrees warmer in the shed than it is outside, so 110F+ in the summer. I'm guessing that the ones that have been out there for 5 years, both SSD and spinning platter, are probably toast now.
“Common sense is not so common.” — Voltaire
In an enterprise environment the limiting factor isn't usually shelf life of the drives.
I work for a company selling enterprise storage, and we expect our customers to use the entire write endurance of an SSD in three years. The reason is that SSD is still quite expensive per terabyte, so it's only used for workloads which really need it. That means database indexes, file caches etc. These generally involve high data turnover and even with wear levelling the whole lifetime will be used up in 3 years.
Lost another SSD over the weekend. Crucial m4 512GB. Lost detection of the drive by the computer (Win 7 desktop), plugged it in through a USB adaptor and it's still not detected (Windows and Mac). That's 3 in the last 18 months.
RIP
Muskin Chronos 120GB (Windows 7 laptop)
Crucial m4 512GB (MacBook Pro 2012)
Crucial m4 512GB (Windows 7 desktop)
That being said I run everything on SSD: 2 HTPC, 2 desktops, 2 MacBooks, 2 Windows laptop.
I can't find the common factor that causes the failures. It would just be working one day, then next day fail detection by the computer and it's all gone.
The problem is hard drives have different properties that aren't understood by pretty much everyone, including hardcore nerds.
SSD's have not been around long enough to actually say they will last X amount of time because even the least-reliable TLC based SSD's have only been around less than a decade. Flash still works by "holding a charge" despite what kind of nonsense gets thrown about. It's not like a CD-R/DVD-R where there are holes burned into a dye substrate and the data will last forever if you can put the discs in a climate controlled environment. Even a regular hard drive will lose data over time, if not the heads seizing due to poor climate control.
That said, there are entire hard drive batches that fail in data centers, often within days of each other. That is due to defects in engineering. If one drive dies, you then quickly run diagnostics on every drive you are using that came from that batch and cycle them out if they are reporting the same pre-fail data as the first one. I've experienced this first hand and had to reinstall the OS and software on servers because they were bought at the same time, so not only did the primary fail, the backup fail-over also failed before the main could be restored.
Even though this sounds reassuring, I started creating par2 checksums for my family pictures (and then back up the whole bunch, of course).
If you run OS X on the desktop, it installs nicely via Homebrew:
$ brew install par2
Then use as follows:
$ cd familypics
$ par2create par2file *
And to verify:
$ cd familypics
$ par2verify par2file.par2
It takes about 5% of extra storage. If you run Linux, you can get that back by using btrfs and mounting it compressed.
8 of 13 people found this answer helpful. Did you?
I wouldn't be surprised if we see a slew of products in the future where you're only licensed to use them until the warranty expires, first sale doctrine be damned. That, and the inclusion of DRM to disable them as soon as the warranty^H^H^H^H^H^H^H^Hsubscription expires.
The drives will instantly brick themselves at the following Monday.
That's being optimistic. More likely is Saturday night at 3AM, resulting in a call in to repair the downed system. Bonus points if you don't require people to be "on-call" rotation, so you have to find someone that isn't too drunk to come in.
It is coincidental that this article comes out just as I lost a lot of data on my computer last night after shutting it off. I have an SSD, however I think it might have been because I didn't let Windows 8 shut down cleanly I simply powered it off. I can't be sure if it is because of the unclean shutdown or being off for a number of hours. When I booted my machine just moments ago and tried to login, it said the nvidia and intel graphics drivers were corrupt. Also, my laptop's updater tool is missing from the program files. I don't know how many different files I might have lost. I was able to boot into safe mode and start repairing, however. Note: I don't normally just shut it off like that, but I was heading to bed and I was bit groggy so I just held the power button out of apathy. I really didn't expect an utter catastrophe to follow.
You haven't been supposed to just force power off since at least Windows 95, so almost 20 years. You must have been running on battery, normally I go start->shutdown, then walk away and let it do it's thing (or just leave it running). What make/ model SSD do you have?
I shutdown my gaming PC all the time, sometimes for weeks or even months at a time (if I get into a console game instead). I've never lost data on it except when the painter left the heater on and then proceeded to open all the doors and windows downstairs while simultaneously closing my office door. When I got home, it was like an oven in there, and the SSD lost all data, but when I reformatted and reinstalled it it continues to work to this day.
Peter predicted that you would "deliberately forget" creation 2000 years ago...
Although Alan Cox probably has something to say about SSDs, I don't think anyone bothered to contact him for this article.
How many fifteen year old laptop drives have you tried to spin up lately? Now compare that failure rate with optical or tapes.
As I wrote above, the life has been extended (it used to be a lot of failures in a lot less than fifteen years) but the underlying modes of failure like the one I mentioned above are still there. Another unpowered failure mode is highly polished parts diffusing together over time so the drive "sticks", once again less of an issue than it used to be due to design changes taking that into account - quite a lot of it done around 2000.
You mean you've never repaired a fileserver while drunk?
Well designed SSDs are designed such that they can handle a sudden loss of power. Unfortunately many of the consumer-grade SSDs being sold right now lack any form of power loss protection and can loose data if they aren't given enough time to power down properly (that includes many SSDs being sold by OEMs pre-installed in desktops and laptops).
When buying an SSD, I always check reviews and the better reviews detail the power loss protection (or lack thereof) present in the drives they review.
Drives with large capacitor banks and write-through caching that can withstand and form of power loss often only cost slightly more than comparable drives without such features. It is well worth it to thoroughly research any SSD before buying and ensure it can withstand sudden power losses without any possibility of loosing or corrupting critical data such as internal page tables and block metadata.