SSDs Fail. Period.

ardichokeardichoke Icrontian
edited June 2011 in Hardware
I've taken a lot of flack from some of you since I started posting on Icrontic for my outspoken criticism of SSDs and their reliability. Most of my criticism has been rooted in my personal experience at my job where I have seen an inordinately high percentage of SSDs die as compared to traditional SATA and SAS drives. I've been told by at least one of you that I'm downright wrong about the lack of reliability of the drives. Well... not only am I not wrong, but now even some of the ardent (higher profile) supporters of SSDs are admitting that they're not as reliable as many people like to make them out to be.

http://www.codinghorror.com/blog/2011/05/the-hot-crazy-solid-state-drive-scale.html

Now then, am I saying that nobody should use SSDs? Surely not. If you really need better disk performance, an SSD is most likely the way to go. You damn well better make sure that you have backups though (as well as the spare cash to replace the drive) because, no matter what Thrax says, they fail... and they do it just as often if not more often than traditional drives. Plus when a traditional drive fails you can often recover some if not all of your data. When an SSD fails, you can kiss your data goodbye.
«13

Comments

  • shwaipshwaip bluffin' with my muffin Icrontian
    edited May 2011
    good thing the plural of anecdote is data.
  • ardichokeardichoke Icrontian
    edited May 2011
    How much anecdotal evidence does it take before it becomes data? I work for a company that has thousands of severs, hundreds of which have SSDs and we've seen a much higher incident of failure among the SSDs than the traditional drives. This spans multiple brands of drives too. That enough data for you? Do I have to manufacture the drives for my numbers to matter? I mean, if anything I would think that drive manufacturers numbers would be the ones called into question since, you know, they're trying to SELL the damn things.
  • shwaipshwaip bluffin' with my muffin Icrontian
    edited May 2011
    I'm saying that you're not giving us that data. You're saying "oh, a bunch of drives fail at the place I work". The blog is saying "oh, a bunch of drives failed for this one guy".

    Give us something like what percent of drives fail in 6mo/1yr/2yr/etc, broken down by manufacturer (maybe). If it is consistently high across all manufacturers or you can say that between the 1 and 2 year mark, a lot of drives die.

    And then compare that to the percent of mechanical drives that die at those same times. If you want to be fancy, put confidence intervals on those percentages and see if the difference is statistically significant.
  • BuddyJBuddyJ Dept. of Propaganda OKC Icrontian
    edited May 2011
    LOL BUT Y U MAD?
  • ardichokeardichoke Icrontian
    edited May 2011
    I don't have access to the precise numbers (at least not precise enough ones to satisfy you SSD fanboys) and even if I did I would probably face disciplinary action if I released them. The best I'm able to say is that there is a disproportionately higher failure rate of SSDs to mechanical SATA/SAS drives.
  • Sledgehammer70Sledgehammer70 California Icrontian
    edited May 2011
    I would love to know if it is one brand or one series of drives... I have had many SSD's running in my systems and have yet to have the issue I have seen with platter drives. Now thats my personal Exp but even system I build for other people seem to not have issues.

    Now your comment about data on a failed SSD is true. I typically have my OS and main programs running on SSD's but the valuable data is saved on a data (Platter drive) and is backed up weekly.
  • ardichokeardichoke Icrontian
    edited May 2011
    I'm not comfortable posting any more information relating to my work at this time. I'd really rather not risk losing my source of income especially since MI is a "right to work" state, meaning I could be fired for no reason at any time. The types of SSDs that we sell are listed on our site. I was just pointing out that I'm not the only one that has problems with SSD failures, in fact some well-known techie SSD enthusiasts have had multiple problems with them as well.

    @shwaip - that blog is not just one guy. It's one guy saying 2 of the 3 he purchased failed, and another tech savvy fellow he knew had his only one fail, and a 2nd tech savvy fellow that he knew had 8 of various makes and models all fail.
  • BuddyJBuddyJ Dept. of Propaganda OKC Icrontian
    edited May 2011
    Oh, good to see the fanboy card has already been played. It's best to throw it down early in a discussion so people know you're not biased or anything. ;)

    I'd LOVE to see real data on SSD failure rates and the potential causes. Shwaip's suggestion would do nicely.
  • NiGHTSNiGHTS San Diego Icrontian
    edited May 2011
    Well, this should end well.

    I don't have a horse in this race, but for full disclosure I'll say I have an SSD running in my rig right now (with the presumption that it would end-of-life well after I'd moved on to something else anyway). In any case, I'll second the fact it's hard to believe anything one way or another without hard data to stare at. I'd really like looking at your particular company's failure rate of SSDs, if that's possible. Until then, I'll have to go off what (a rather large pool of internet users more knowledgeable than I with this technology) are telling me: they're good, they last, and they're fast.
  • CantiCanti =/= smalltime http://www.youtube.com/watch?v=y9K18CGEeiI&feature=related Icrontian
    edited May 2011
    What is this the new Nintendo handheld? Super Saiyan Dual Screen?
  • BasilBasil Nubcaek England Icrontian
    edited May 2011
    This thread requires moar stats.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited May 2011
    Waiting for real data.
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited May 2011
    If this was a significant enough problem, I would see more google results past 2008 than just the blog you linked. /shrug
  • ardichokeardichoke Icrontian
    edited May 2011
    Thrax wrote:
    Waiting for real data.

    I'm not risking getting fired to satisfy you, since even if I gave you real data you'd just regurgitate industry supplied stats at me and tell me I'm wrong.

    All I'm saying is that a more influential and trusted blogger than you, who was (and still is) a proponent of SSDs has admitted that they fail, they fail often and when they fail they do so spectacularly. Regardless of the fact that his sample size is only 12 SSDs, it still puts a dent in your "omg, SSDs are the greatest thing ever, they can sustain 20GB of writes per day for 20 years always blahblahblah" that you post everywhere.

    Also, as I said, I'm not telling people NOT to use SSDs. I'm just pointing out more information backing up what I've been saying for the past 2 years. SSDs are not as reliable as the Thraxes of world like to hype them to be... so make sure your data is backed up and you have a spare laying around if your computer is mission-critical.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited May 2011
    ardichoke wrote:
    All I'm saying is that a more influential and trusted blogger than you.

    Not data.
  • ardichokeardichoke Icrontian
    edited May 2011
    Thrax wrote:
    Not data.

    Where's YOUR data. I'm talking the stuff that's not industry supplied, I want first hand data that backs up the reliability claims you've made.
  • CantiCanti =/= smalltime http://www.youtube.com/watch?v=y9K18CGEeiI&feature=related Icrontian
    edited May 2011
    Thrax wrote:
    Waiting for real data.

    2zsywzb.jpg

    This chart provided by 10 hours of research in MS paint through a nonprofit, unbiased organization.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited May 2011
    ardichoke wrote:
    Where's YOUR data. I'm talking the stuff that's not industry supplied, I want first hand data that backs up the reliability claims you've made.

    The burden of proof is on you.
  • ardichokeardichoke Icrontian
    edited May 2011
    Thrax wrote:
    The burden of proof is on you.

    Why? Because I'm advocating that people backup their data? You're the one that's been making ridiculous claims about SSD reliability.
  • ardichokeardichoke Icrontian
    edited May 2011
    Canti wrote:

    This chart provided by 10 hours of research in MS paint through a nonprofit, unbiased organization.

    My rage is entirely based on people that lose their data because they don't have backups of it, their SSD fails, then they throw a fit at me because they have unrealistic expectations of SSD reliability thanks to people like Thrax overhyping the technology.

    People like Thrax give people unrealistic, industry supplied numbers which instill unrealistic expectations which then make the lives of people that have to deal with the consequences miserable. I hate them.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited May 2011
    Why? Because I'm advocating that people backup their data?

    Cut the shit, brohammer. Don't try and revisit an old and, on my part, completely forgotten debate about the reliability of flash memory and then change the nature of the debate when I push you for something more than anecdotal evidence.

    You can do better, and I expect better.

    Also, I will do exactly fuck all about this topic when I'm facing unprovoked posturing like this even before I take a seat at the table:
    All I'm saying is that a more influential and trusted blogger than you
    because, no matter what Thrax says
    I'm not risking getting fired to satisfy you, since even if I gave you real data you'd just regurgitate industry supplied stats at me and tell me I'm wrong.
    "omg, SSDs are the greatest thing ever, they can sustain 20GB of writes per day for 20 years always blahblahblah"
    they have unrealistic expectations of SSD reliability thanks to people like Thrax overhyping the technology.

    You can sit there all night long and ragepound the keyboard like an angry fucking chimp, but I won't play ball with it. Why should I? What do I have to gain by enduring your arrogant tantrums?

    Nothing.

    Find someone else to care.
  • MyrmidonMyrmidon Baron von Puttenham California Icrontian
    edited May 2011
    ardichoke wrote:
    How much anecdotal evidence does it take before it becomes data?

    ACHIEVEMENT UNLOCKED: Answer a loaded question originally posed to sympathetically strengthen an argument.

    http://www.osra.org/itlpj/bartlettkotrlikhiggins.pdf

    Quick journal searches (and yes, google scholar searches too... I'm not too proud) reveal no statistical studies simulating the reliability of SSDs, most data at this point is for specific variables (magnetic susceptibility, humidity, temperature, speedy CTE changes, etc). Overall reliability data would be a theoretical construct out of these calculations along the lines of 'which one gonna fail first?' Done by - guess who cares about the reliability of the device - industry.

    Calling them "industry statistics" while hinting about their inaccuracy is a disservice. It's a disservice to the scientific studies behind them, and it's a disservice to the engineer that came up with the factor of safety for that number. I used to work in an industry that built numbers like that.

    The only people who will do reliability research are the people that care. University researchers will be much more interested in performance. University research on reliability will be (more often than not) funded by industry grants (as opposed to other third party organizations), making even these studies "industry statistics." I used to work in a university lab that performed "industry" studies like that.

    A better request would be "In order to refute a claim that I am bringing against established claims, I want YOU to show me some first-hand, impossibly extensive research that hasn't yet been done and therefore must be self funded. Any study that at one point was done using existing research systems will be discarded because I prefer my research photodocumented on websites rather than in a scholarly journal."
  • shwaipshwaip bluffin' with my muffin Icrontian
    edited May 2011
    Ardi, the burden of proof is on you. You're making the claim that SSDs are significantly less reliable than existing mechanical disk technology.

    We all understand why you don't want to publish data from your job. It would be very interesting data to see, but not at the price of your income. However, if you cited sources other than what amounts to a guy with a blog and two other people he knows (however well respected that blog may be), I'm sure you'd get a lot less flak.

    Personally, I wouldn't be surprised if SSDs were more prone to failure - they're a newer technology...however until I see something detailed like this paper (PDF!), you have to understand why I and others would be skeptical.
  • ardichokeardichoke Icrontian
    edited May 2011
    First, I'm not your "brohammer" so drop that crap. I don't wear popped collars. I don't GTL. I don't have any patience for airheads.

    I just happened to see this story while watching some Tekzilla today and thought I'd bring it up. Sue me if I preemptively mentioned you since I knew you'd post the same. old. shit you post every time I tell someone that plans on using an SSD to back up their data because they're not as reliable as the industry likes to claim.
  • ShortyShorty Manchester, UK Icrontian
    edited May 2011
    "Stuff" fails. Platters fail. Tiny itty bitty electronics fail. It's what keeps us in a job/occupied/exacerbated. Stats are are never really representative.
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited May 2011
    Thrax tried to drop that reasonably and Ardi's ragepants won the day. Good times in the fast lane.
  • KwitkoKwitko Sheriff of Banning (Retired) By the thing near the stuff Icrontian
    edited May 2011
    On a long enough time line, the survival rate for everything drops to zero.
  • BuddyJBuddyJ Dept. of Propaganda OKC Icrontian
    edited May 2011
    tl:dr summary

    Back up your stuff regularly to minimize risk of total data loss.
  • TushonTushon I'm scared, Coach Alexandria, VA Icrontian
    edited May 2011
    Buddy J wrote:
    tl:dr summary

    Back up your stuff regularly to minimize risk of total data loss.
    I think your avatar summarizes this whole thread.
  • _k_k P-Town, Texas Icrontian
    edited May 2011
    Am I too late?

    427517_capt.b248c3cd64c74786b7aad27b173f337a.aptopix_obama_2008_orjh118.jpg
Sign In or Register to comment.