7 Replies Latest reply: Jun 21, 2010 7:52 PM by bdwilcox RSS

SSD data retention for archived drives

bdwilcox Community Member
Currently Being Moderated

Hi,

 

I work in litigation support for a large corporation and have been trying to warn them that the data retention time for an SSD is shorter than the 10 years of so for standard physical drives.  They archive users' physical hard drives and count on the data on them being available almost indefinitely.  Does Intel or anyone have a white paper or info I can use to report to our IT people here about data longevity issues with SSDs that are sitting in storage and not installed in a PC?

  • 1. Re: SSD data retention for archived drives
    NandFlashGuy Community Member
    Currently Being Moderated

    Hi bdwilcox,

     

    The official method of measuring data retention for a non-volatile memory is published by JEDEC (a semiconductor industry standards group).  The specification for data retention is JESD22-A117B.  You can download it from here.

     

    The actual data retention time of a flash memory will depend upon many factors:  the quality of the Nand flash, the number of program/erase cycles, the rate at which these program/erase cycles occured, the temperature at which the Nand flash will be stored at, etc.  Depending upon these factors, the data retention may exceed 10 years or be far less.

     

    Ignoring the data retention aspect, I would argue that cost/benefit for a high-performance SSD is more compelling for active client use compared versus data archival.

     

    Hope that helps.

  • 2. Re: SSD data retention for archived drives
    bdwilcox Community Member
    Currently Being Moderated

    Hi NandFlashGuy,

    Thank you for your answers and the JEDEC spec. They will definitely help me in understanding NAND data retention behavior.

     

    I  realize that any answers to my next couple of questions probably wouldn't have any real data to back them up, but I was wondering if you would share your educated opinion about them.

    1) Do you think the 10 year retention time of NAND may be lengthened or shortened once it's soldered onto an SSD’s PCB?

    2) Does leakage rate vary by NAND type (SLC vs MLC)?

    3) Have you seen any studies of whether dormancy makes NAND cells more prone to failure or data loss.  i.e. when I access an SSD that’s been sitting in storage for five years, will the sudden introduction of current affect the module more than if it had been used all along.

    4) Our drives sit in an old salt mine underground, so cosmic rays aren't much of a concern. What worries me more is background radiation from the mine itself.  Do you have any knowledge of whether long-term, low-level background radiation can effect NAND or SSDs?

    5) We use some nasty research magnets in the labs that would wipe a standard hard drive in the blink of an eye.  Though it might seem that SSDs wouldn’t be prone to data loss from these research magnets, do you have any knowledge of whether extreme magnetic fields (MRI level) can effect data on an SSD?  Could a strong enough magnetic field possibly induce a current in the traces of an SSD that might flip gates?

    Again, I realize they're just guesses on your part but any insight would be much appreciated.

    Thanks again.

  • 3. Re: SSD data retention for archived drives
    NandFlashGuy Community Member
    Currently Being Moderated

    Hi bdwilcox,

     

    You asked a bunch of questions:

     

    1) Do you think the 10 year retention time of NAND may be lengthened or shortened once it's soldered onto an SSD’s PCB?

     

    Given that data will be programmed into the Nand after the solder process, there should no change in data retention capabilities.

    2) Does leakage rate vary by NAND type (SLC vs MLC)?

     

    At an equivalent amount of cycling, SLC will normally have longer data retention than MLC.  Equivalently, for the same data retention requirement, SLC can survive more program/erase cycles compared to MLC.  But for a typical PC user, MLC should be more than sufficient.  Enterprise servers are using SLC flash though.

     

    3) Have you seen any studies of whether dormancy makes NAND cells more prone to failure or data loss.  i.e. when I access an SSD that’s been sitting in storage for five years, will the sudden introduction of current affect the module more than if it had been used all along.

     

    I'm not aware of any studies like you mention, but data retention measurements are done in the "dormant" state (units are placed into an oven).  The sudden introduction of current should cause no problems.  The temperature at which the Nand is stored can have a dramatic influence on the data retention capability.

     

    However, by keeping a drive in active use, there will be a natural "refreshing" of the data on the SSD.

    4) Our drives sit in an old salt mine underground, so cosmic rays aren't much of a concern. What worries me more is background radiation from the mine itself.  Do you have any knowledge of whether long-term, low-level background radiation can effect NAND or SSDs?


    I'm not aware of the effects of low-level background radiation on NAND.  I imagine the basics aren't dramatically different from any other semiconductor -- some rate of soft errors may occur.  However, Nand specifies some amount of error correction as a requirement, so that gives Nand some margin for these types of effects vs a situation where every single bit must be perfect (like normal DRAM vs ECC DRAM in servers).  There are uses that require "military grade" semiconductors / packaging, but typically this is more expensive due to the relatively small demand.

     

    5) We use some nasty research magnets in the labs that would wipe a standard hard drive in the blink of an eye.  Though it might seem that SSDs wouldn’t be prone to data loss from these research magnets, do you have any knowledge of whether extreme magnetic fields (MRI level) can effect data on an SSD?  Could a strong enough magnetic field possibly induce a current in the traces of an SSD that might flip gates?

    Flash works by trapping electrons on a floating gate of a transistor.  So first order, it should be robust against normal magnets.  Extreme magnetic fields may have second order effects.  Just taking a guess, but I would wonder about the whether an extreme magnetic field requires higher quality solder/board reliability.  But I don't think you have to worry about flipping gates of either the Nand or the controller of the SSD.

     

    You may have better luck searching the IEEE journals for more specifics about your particular situation.

  • 4. Re: SSD data retention for archived drives
    bdwilcox Community Member
    Currently Being Moderated

    Wow, NandFlashGuy, I can't thank you enough for the insight you've given me.  I feel A LOT more comfortable with our company's eventual transition to SSDs.  I really appreciate your taking the time to answer these admittedly esoteric questions and hope it might give others insight into how SSDs fit into their records management strategy. It was a real pleasure conversing with you and thank you again.

  • 5. Re: SSD data retention for archived drives
    bdwilcox Community Member
    Currently Being Moderated

    NandFlashGuy, one quick follow-up question.  You stated that temperature will affect data retention in NAND chips.  I've looked at as much IEEE material as I could find and the only thing I could find regarding NAND and temperature is that working NAND chips are negatively affected by higher temperatures.  What I couldn't find was reference to temperature impact on NAND data retention in idle chips. Do you have any further info or reference you could point me to?  Are we talking temperature extremes or those within human tolerance (0-105 degrees F)? In other words, would hard drives stored at 35 degrees F show measurable data retention differences to those stored at 90 degrees F?  Thanks for any additional info you can provide.

  • 6. Re: SSD data retention for archived drives
    NandFlashGuy Community Member
    Currently Being Moderated

    Hi bdwilcox,

     

    Yes, there may be a measureable difference between 90C and 30C at a high enough cycle count.

     

    In fact, JEDEC specifies the method of modeling long term data retention at room temperature via shorter term data retention at higher temperture.  The relationship is related to the activation energy and is modeled by the Arrhenious equation.

     

    Here's a link to a Freescale datasheet (they used to partner with a NOR flash vendor) that I found via a quick search:

    http://www.freescale.com/files/microcontrollers/doc/eng_bulletin/EB618.pdf

     

    In the Freescale example:

    • 1008 hours at 150C is equivalent to 1150 years at 25C

     

    Of course the specifics will vary depending on the Flash, but this at least gives you some decent first-order estimations.

  • 7. Re: SSD data retention for archived drives
    bdwilcox Community Member
    Currently Being Moderated

    NandFlashGuy, thank you again for the first rate info and sorry to bother you again with that extra question.  BTW, the difference in that Freescale example is simply staggering.  This is definitely something I will need to look into regarding our current storage environment.  Much thanks. -bdwilcox

More Like This

  • Retrieving data ...

Legend

  • Correct Answers - 4 points
  • Helpful Answers - 2 points