Solid-state drives lose data if left without power for just a few days

Discussion in 'hardware' started by Minimalist, May 10, 2015.

  1. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    Got spinners that are 32+ years old which are still operational. I don't believe current consumer grade solids will last as long.
     
  2. NGRhodes

    NGRhodes Registered Member

    Joined:
    Jun 23, 2003
    Posts:
    2,381
    Location:
    West Yorkshire, UK
    Problem is its very unpredictable with HDDs. Some last way beyond their warranty period, then some don't even last 6 months, you can't predict and have to have a redundancy/backup/replacement plan if your data is critical.
     
  3. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    And no doubt, today's motherboards don't come with interfaces to read 32 year old HDs.
     
  4. Rasheed187

    Rasheed187 Registered Member

    Joined:
    Jul 10, 2004
    Posts:
    17,546
    Location:
    The Netherlands
    OK, then I misunderstood. And I guess it's a good thing that I almost never turn off my PC (with SSD), it's in standby-mode most of the time.
     
  5. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Do you keep your house at 86°F? Or closer to 70 - 75°F and maybe closer to 68°F in the winter?

    30°C (86°F) for inside your living or office space is hot. Most computer users don't live in such environments without AC, or they have those temps for only a few days out of the year.
     
  6. Rasheed187

    Rasheed187 Registered Member

    Joined:
    Jul 10, 2004
    Posts:
    17,546
    Location:
    The Netherlands
    AFAIK, most of the time it's about 20 to 25 degrees. In the summer (with no AC) it's often 30 to 35 degrees.
     
  7. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Still, that is only during the day, right? And you use the computer regularly, so no worries.
     
  8. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    But cheap adapters can be had for $10. Or interface directly with the original host hardware. It would be prudent to keep that capability alongside with the hard disk.
     
  9. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    Data backups are important regardless of what media you use.

    Anyone doing long-term storage on standard HDD's will eventually come across the issue of data migration. They'll even figure out the need for migration themselves as time goes on. Performance and capacity increases, new formfactors & interfaces.. All enough to prompt someone to switch to the newest media.

    Furthermore, these standards change slowly enough that even the earliest consumer drives from the 1980's can still be read. That's a 30 year window!

    I have data from the mid 1970's still archived and it sits on modern media.
     
  10. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Yep! And multiple backups too - preferably with at least one off-site.

    A friend thought I was being over paranoid by keeping a backup of all my data on a drive in the safe deposit box at my bank. I kept telling him to keep a backup in his bank, or at family or friend's house but no. He kept all his backups to all his computers on attached drives on his desk. Worked great until a badguy broke into his house and stole his TVs, home stereos, wife's jewelry, plus all his computers and his attached backup drives. He had no way of recovering anything. :( And of course, insurance only covered the hardware, not the data. And for many, the data is worth MUCH MORE than the hardware. It is for me.

    So keep an off-site backup - not just in case of badguys breaking into your home, but fire, floods, and lightning strikes too.
     
  11. xxJackxx

    xxJackxx Registered Member

    Joined:
    Oct 23, 2008
    Posts:
    8,625
    Location:
    USA
    The article is sensationalistic click-bait. The title implies that if you leave your computer off for the weekend (or something similar) your data will be gone when you turn it back on. I've had SSDs sit for months without issue.
     
  12. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
  13. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    Unofficially, Flash engineers suggest a 10year lifespan of bios chips a smaller capacity utility roms. Lifespan of data on tlc flash is perhaps 5 years at best. Power on or off.. Irrelevant.
     
  14. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Oh? Got a link? I'm seeing reports that suggest this depends on the usage expected lifespans well above that, up to 70 years even.
     
  15. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    Check some data sheets for the individual parts. I've also had this discussion with some engineers.
     
    Last edited: May 18, 2015
  16. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    SLC can theoretically go above 50 years, cell size, charge, all that stuff, it's all good for SLC. MLC less so. And TLC pretty bad. TLC is the budget stuff of the Flash industry.

    It's a good thing the 3D stacked NAND has to back up and use a larger geometry, inherently giving better life.

    In fact Apple is having problems erratic operation of TLC NANDs used in some iPhone 6+ variants. Apple knows about this. And the accepted solution is to keep returning the phone till you get one that is MLC based.
     
  17. quietman

    quietman Registered Member

    Joined:
    Dec 27, 2014
    Posts:
    511
    Location:
    Earth .... occasionally
    I'm now convinced that the title of this thread is more alarmist than factual ....
    .... and the title is merely a quote from the article on zdnet.

    Of the many SSD devices that I have , only one has ever failed.
    It was a 2 GB USB stick , over 10 years old , and no named manufacturer .... a promotional give-away.
    And even then , it wasn't a case of data loss / corruption .... just a total failure of the device.

    I've yet to hear of a single instance of actual data loss / corruption.

    I know there are various postings on here along the lines of " USB device not seen / recognised " ... etc.
    .... but that is a completely separate issue.

    BTW , thanks Bill_Bright for the link to an interesting article ..... #37
     
    Last edited: May 18, 2015
  18. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    I have checked data sheets and I have had discussions with fellow engineers too. I provided a link to several "current" reports that do not support your claims. And I provided supporting evidence because I don't expect you, or others to just believe me because I say so. Please reciprocate in kind.

    So I ask again, got a link? And I would ask it be to something current.

    Yeah, we're still waiting for that. And unfortunately, even if someone makes such a claim, without knowing all the circumstances, it really would be impossible to verify any loss of data integrity was actually cause by the lack of data retention during long term storage.

    Was the data verified intact before put in storage?
    Was the device properly stored in it's plastic case the entire time?
    Did the handler take the essential ESD precautions before, during and after storage?
    Was the device left undisturbed throughout the many years in storage?
    What were the actual environmental conditions (temperature, humidity, exposure to light) while in storage?​

    Because there are many things that can cause the loss of data in these devices, these claims need to be validated in a controlled laboratory test environment. One or two individual "normal" users claiming the data was intact before storage, and now years later, the data is not, is NOT proof data integrity was lost exclusively due to loss of retention over time.

    Let's not blame Minimalist (the OP of this thread). Again, I blame Zack Whittaker and ZDNet for carelessly, ineptly and incompetently misrepresenting the facts with his false claim of "new" research. There is nothing new about it.

    If anything, I thank Minimalist for starting this tread because it allowed us to realize the truth.
     
  19. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&uact=8&ved=0CDIQFjAA&url=http://users.ece.cmu.edu/~omutlu/pub/flash-memory-data-retention_hpca15.pdf&ei=gzNaVbD2H4WSyAS6-YD4Dg&usg=AFQjCNHJSDQwLFDGBq_D5W_iSuv04dPDgA&sig2=_HlgkjHS5oDcLTW1yXKNMg

    After 28 days with no power, the error correction algorithms need to change significantly and adapt to the degrading cells. We put trust in that that the engineers got that algorithm right. Fully 56% of the work a NAND controller does is related to fixing errors on the fly to present you with error free data. You can thank that to TLC and QLC for a worsening trend here.

    Considering the precariousness of the data stored in these devices and the amount of ECC needed I wouldn't trust them for any archival storage. I suppose it's time for someone to run a test where they take a new device and fill it with data and see how long it lasts. Put it in a box and revisit it 5 years.

    Most new parts are spec'd at 100 years retention. As soon as you put a handful of reads and writes that drops to 10 years. Reading affects the strength of the charge in the cell, many reads to one cell will cause the controller to refresh as the charge weakens. This indirectly shortens the life of that cell.
     
    Last edited: May 18, 2015
  20. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    Ohh wait. I have a couple of CF cards from the 1990's I haven't touched since. They are from an old digital camera. I put it in storage with my classic Canon equipment I keep for sentimental reasons. These memory cards would be from 1996. I'm willing to bet they are still ok and that whatever is on them is alright.

    I have a USB JumpDrive that I hadn't powered up all year. It's been sub-zero in the car trunk, and is now roasting in there. I bet it's alright.

    What about new PC's? Surely these come with SSD pre-loaded with and OS and shovelware. According to ziff davis guy, these shouldn't work. Surely they are going to be exposed to temperature extremes in transportation and from the time they leave the factory burn-in test to the time you open the box on christmas morning more than 30-days may have elapsed..!

    I hate sensationalism. And reading the comments further muddies the waters. It's like a goddamned circus trying to find a concrete answer.

    IDK. I'll just stick to what works. A traditional old-school backup plan.
     
  21. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Thanks for that link! :) That was certainly a comprehensive study, and I was particularly pleased to see the study was done at a more realistic 20°C (68°F).

    But I am afraid it really does not answer the concerns because of the many variables involved. It clearly shows that the charge potentials do start to degrade rather quickly, but it does not show how long or far it can degrade before the inherent error correction routines fail to recover/correct for any "data" retention problems. In other words, it illustrates that "charge retention" is not the same as "data retention" because a cell can lose much of its charge, but still maintain the data. And it does not show how long a SSD can be in storage without power before charge retention degrades so far that the data is no longer retained, or recoverable through normal error correction.

    But I think it does illustrate what we've been saying along and what I said over a week ago, SSDs were never intended for long term storage without power. That is, SSDs are not for archival or backup purposes. For that, with normal consumers, the HD still rules.
     
  22. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    That's an excellent point! I note more and more notebooks and other devices are coming with SSDs. And many are shipped with the batteries disconnected and remain that why for extended periods in warehouses and on store shelves until purchased. And while some warehouses are cooled, I don't know of any warehouse that is air conditioned to 68°F except for perishable foods, medicines and such.
     
  23. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    This is a complex topic if you discuss it on a technical level. And that leaves room for confusion when it's digested for the layperson. Therefore I believe the industry will spin it whatever way results in the most sales.

    With this kind of ambiguity I'll just stay with the time tested procedures and methods which have served me so well over the years.

    I also believe many tech articles regarding SSD longevity are good only for triggering a discussion, and raise more questions then they answer. Or don't answer in a concise enough manner.

    Much of the uncertainty comes from the author not knowing the charge capacity of a cell inside an arbitrary flash chip. And from not knowing the details of ECC. Remember,the SSD spends more than 50% of the time fixing inherent errors. A typical HDD may do 10% in comparison. That's on every single read request.

    A typical archival system would ideally do 0%. But in reality they manage as low as 2 or 3 percent.
     
    Last edited: May 18, 2015
  24. Keatah

    Keatah Registered Member

    Joined:
    Jan 13, 2011
    Posts:
    1,029
    These routines vary wildly, they're determined by the firmware, the controller, the layout of the NAND array, speed vs reliability, cost, how aggressive the wear leveling is, the intended target audience. And more. To make matters worse, they go through many revisions for each product line of any one arbitrary brand. These ECC routines are proprietary. And without seeing how they work inside I believe any reports can only summarize on the big picture. In different terms; treat them as black boxes and develop your operational procedures accordingly.

    I also believe that makers of Spinner disks are going to have to start touting permanence and reliability once SSD sales exceed some yet to be determined pain point for them. Then we'll get another load of debates.

    It is important to understand that no SSD cannot function without these routines. The flash array is constantly in error if you compare what goes into it and what comes out of it. Even if it is brand spank'n new.
     
  25. Bill_Bright

    Bill_Bright Registered Member

    Joined:
    Jun 29, 2007
    Posts:
    4,041
    Location:
    Nebraska, USA
    Right - hence my remark that the article/study does not answer our concerns - just too many variables.

    Bottom line for me and my builds is that I will continue to use SSDs on all my builds for the OS, drivers, apps and my user-created data files (Word docs, Excel spreadsheets, emails, etc.). And I will continue to use HDs to backup my SSDs, for my tunes (I've copied my 600+ music CDs to HD) and videos. Plus I will continue to backup everything again to my old XP box I have repurposed as my NAS (it is located in a different part of the house, and has been blocked for Internet access in my router).
     
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.