[ale] SSD info you need to know ( was OT SSD remaining lifetime indicator)

[Ed Cashin]

I agree that it was very interesting.  Ron Frazier, if you remember
the sources you used, would you mind please listing them?  I like to
keep track of sources when I find good info.  If it's too much
trouble, nevermind and thanks anyway!

On Sat, Sep 8, 2012 at 3:36 AM, Sam Rakowski <devnull at iamdevnull.info> wrote:
> On Sat, 8 Sep 2012, Ron Frazier (ALE) wrote:
>
>> Date: Sat, 08 Sep 2012 01:56:35 -0400
>> From: "Ron Frazier (ALE)" <atllinuxenthinfo at techstarship.com>
>> Reply-To: Atlanta Linux Enthusiasts <ale at ale.org>
>> To: Atlanta Linux Enthusiasts <ale at ale.org>
>> Subject: Re: [ale] SSD info you need to know ( was OT SSD remaining lifetime
>>     indicator)
>
>> Hi guys,
>> I've been doing more research into SSD's since Rich posted the info about that special that Tiger Direct had.  This topic is complex and confusing.  It's late at night and I'm tired, but I wanted to post several things I've found out.  I hope to post more in depth data later, maybe.  Finding data on SSD reliability is hard and there's lots of FUD (fear uncertainty and doubt) out there.  Here is a synopsis of what I believe to be reliable data.  I don't have time tonight to distill and quote all the sources.
>> Intel seems to be the king of the hill in terms of reliability.
>> Much of this probably also applies to memory sticks and memory cards.
>> 1) NAND endurance shouldn't be a problem with GOOD QUALITY consumer drives in consumer applications when you're writing less than 10 GB / day to the drive.  See the specs and warranty for details.  The drive should last it's warranted time, usually 3 or 5 years.  A good measure to look at is the TBW (terabytes written) spec, which would give you a maximum about of data you can write before degradation.  Use the worst case number.  If the drive provides it, the media wear indicator should provide a reasonable indicator of remaining life percentage, but not necessarily time.
>> 2) Some warranties end when the media wear indicator reaches 1 (starting from 100).  So, if you use it within the specified limits, you get the 5 year (for example) warranty.  If you exceed those limits, in effect, it wears out when it wears out, and your warranty is over.
>> 3) What is more alarming is that SSD's are, in many cases, more susceptible to glitches caused by power failures, system lockups, firmware problems, controller failures, etc.  Failures of this type can sometimes trash all your data.  Some of this you cannot control.  However, you probably want to review published data on fault protection for your existing or proposed drive, make sure it has power fail protection or add some of your own, and upgrade to the latest firmware and upgrade again if you hear about problems that have been corrected.
>> 4) NAND chips made on a smaller fabrication process, 25 nm vs 34 nm, etc., have LESS endurance but have a LOWER price.  Typical endurance for 25 nm MLC NAND is 3000 program / erase cycles.
>> 5) Look for a controller architecture with a typical WRITE AMPLIFICATION of less than 1.  Higher numbers wear out the chips faster.  The Sandforce 2xxx controllers are highly regarded here.
>> 6) I find this one very disturbing.  Many of the drives have a POWER OFF DATA RETENTION spec.  For CONSUMER drives, this is 1 year.  For ENTERPRISE drives, this is 3 MONTHS!  If you want your drive to keep your data, keep it turned on.  Don't put data on the drive and put the drive on a shelf, otherwise you could come back and find the data has vanished.  Supposedly, if the drive is on, the controller is periodically scrubbing the NAND for weak cells and will regenerate or relocate the data.
>> 7) If the drive glitches out, for whatever reason, data recovery may be difficult or impossible.  It will likely be expensive.  You have to find a data recovery company with intimate experience with your particular NAND / controller setup.  Encryption makes things even worse.
>> 8) Although endurance may not be a problem for most users, STUDIES SHOW THAT SSD'S ARE NOT SUBSTANTIALLY MORE RELIABLE THAN HDD'S OVERALL.  Failure rates for SSD's in the first few years of life are in the 1% - 2% range, fairly comparable to hard drives.  In later years of life, HDD's have a failure rate of up to 7%.  Little long term data is available for SSD's.  Half of all HDD failures occur because of non mechanical reasons.  Those non mechanical things are similar in both HDD's and SSD's.  Because of the vulnerability of SSD's to glitches and the difficulty of data recovery, your data may be less safe on an SSD, endurance and physical shock resistance notwithstanding.
>> (I believe that rock solid power protection, both surge protection and power failure protection, is one of the  best ways to improve computer reliability, including the SSD's.  The Smart UPS line from APC can boost sagging line voltage or cut excessive line voltage, to a point.  This type of equipment is especially important in GA, where we have the 2nd highest incidence of lightning in the USA.)
>> 9) I have been able to find very little about the possible failure modes of the drive when it wears out.  However, Intel's documentation recommends that you replace the drive when the media wear indicator reaches 1 to avoid data loss.
>> If you are considering or are using SSD's, I hope you find this useful.  Please feel free to add to the discussion from your own experience.  I certainly found these facts eye opening and somewhat disturbing.
>
> Excellent post. Full of useful information. Thank you very much.
>
> --
> /dev/null
> 4057 0DA0 0983 FFA1 8756  670F 754A 0CB9 A367 275B
> http://devnull.iamdevnull.info/gpg.gpg
> _______________________________________________
> Ale mailing list
> Ale at ale.org
> http://mail.ale.org/mailman/listinfo/ale
> See JOBS, ANNOUNCE and SCHOOLS lists at
> http://mail.ale.org/mailman/listinfo



-- 
  Ed Cashin <ecashin at noserose.net>
  http://noserose.net/e/
  http://www.coraid.com/