MTBF and AFR - how long will drive last?

Discussion in 'SSDs & Data Storage' started by Morlak, Feb 19, 2016.

  1. Morlak

    Morlak n00b

    Messages:
    3
    Joined:
    Feb 19, 2016
    Seagate say that HDD have MTBF of 2M hours. By math, 2M h = cca. 228 years. Does this mean that drive will last 228 years working 0-24h?

    Seagate also say that HDD have AFR of 0.44%. So, if I get this right, for first year is 0.44%, for second 0.88% and so on. By this "half-life" of HDD when half of them will be dead is - 114 years of constant work 24h. My math this is in correlation with upper result, 114 is half of 228.

    To me, 228 years or 114 half-years sound too good to be true.

    So, my questions are:

    1. Am I wrong in my math?
    2. If I am, how can I calculate approximately how long will drive last using known MTBF and AFR from manufacturer?

    Thank you. ;)
     
  2. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,219
    Joined:
    Nov 19, 2008
  3. Morlak

    Morlak n00b

    Messages:
    3
    Joined:
    Feb 19, 2016
    Thanks!

    So, the real formula is R(T) = exp(-T/MTBF). Nice to know, I'm calculating all odds right now. :)


    Just to add, by this formula there is 64% chance that my disk will survive 100 years, and 80% chance that it will survive 50 years. Quite long, perhaps that is the reason why it cost so much.
     
    Last edited: Feb 19, 2016
  4. Brian_B

    Brian_B 2[H]4U

    Messages:
    2,501
    Joined:
    Mar 23, 2012
    That is assuming the MTBF figure released by the manufacturer is to be believed. Not that many hard drive models out there can claim to have survived for 50+ years, so the MTBF figures are, right now, mostly just based on early defect counts and a lot of longevity speculation.
     
  5. Stereodude

    Stereodude 2[H]4U

    Messages:
    3,285
    Joined:
    Oct 20, 2000
    I think it has more to do with HALT/HASS testing.
     
  6. ABLV

    ABLV n00b

    Messages:
    3
    Joined:
    Feb 7, 2016
  7. JoeComp

    JoeComp [H]ard|Gawd

    Messages:
    1,036
    Joined:
    Jan 23, 2012
    Note that a major assumption of the MTBF methodology is that the product has a CONSTANT failure rate.

    If you buy that the actual failure rates for hard drives are the same in the first year as in the fifth year, then I have a bridge in Brooklyn that you might also like to buy.
     
  8. izx

    izx Limp Gawd

    Messages:
    302
    Joined:
    Jul 30, 2007
    Here's the full Seagate disclaimer on MTBF & AFR from the Enterprise series drive manual:

    Note the operating conditions specified. Most importantly, see the bolded/underlined bullet -- the MTBF and AFR do not apply to individual units. You shouldn't rely on them to predict your drive reliability unless you're running at least 10,000 units within the specific conditions. In the first year then, you should budget for 44 of those drives failing.
     
  9. Morlak

    Morlak n00b

    Messages:
    3
    Joined:
    Feb 19, 2016
    What is wrong with the bridges in Brooklyn? :D

    I see now. I wish my calculations were true. :(


    Thank you guys for clearing this matter. ;)
     
  10. izx

    izx Limp Gawd

    Messages:
    302
    Joined:
    Jul 30, 2007
    For reliability purposes, consider other "guaranteed" specs such as the unrecoverable error rate (1 in 10^15 for enterprise, 1 in 10^14 for consumer) and the warranty length (usually 5 years for enterprise, 1-2 for consumer). In general, enterprise drives will have a lower failure rate than consumer drives. If you can afford it, go for the enterprise drives. If you're going for consumer drives, the consensus is that the HGST Desktop NAS series are the best around.
     
  11. zrav

    zrav Limp Gawd

    Messages:
    163
    Joined:
    Sep 22, 2011
    Of course, all studies with statistically significant disk populations point to AFR an order of magnitude higher than what is stated by the manufacturer, except maybe HGST.

    This has been shown to be untrue by backblaze. The advantage of enterprise disks is longer warranty, TLER and constructive measures to reduce vibration in large disk arrays (which impacts seek performance), not lower failure rates.
     
  12. jayhopeishere

    jayhopeishere n00b

    Messages:
    10
    Joined:
    Jun 25, 2013
    I bought 3 HGST 4TB drives and so far 2 of them have gone out. I bought 3 2TB seagates and only 1 of them lasted also. I ran both set of drives in a raid and they were used to record tv shows with WMC. At this point I don't know what brand of drives I can try next. The only drives I never had to replace are some old 2TB Samsung drives that were produced before the flood in Thailand. I regret not buying more than 4.
     
  13. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,219
    Joined:
    Nov 19, 2008
    With an extremely small sample size like this you could just have really bad luck . Remember that someone has to have the ~1% of drives that fail annually for HGST and the 4% or so of drives that fail annually for Seagate. However I would check the voltages on your power supply. It's definitely possible that your environment is destroying the drives.
     
    Last edited: Mar 2, 2016
    izx likes this.
  14. Stereodude

    Stereodude 2[H]4U

    Messages:
    3,285
    Joined:
    Oct 20, 2000
    Do you have a vibration issue? Perhaps a power supply issue?
     
    izx likes this.
  15. izx

    izx Limp Gawd

    Messages:
    302
    Joined:
    Jul 30, 2007
    Or heat? Like prolonged 55+?
     
  16. JoeComp

    JoeComp [H]ard|Gawd

    Messages:
    1,036
    Joined:
    Jan 23, 2012
    How exactly have your drives failed? What are the exact symptoms or indications you are using to determine that they have failed? Have you done any follow-up tests to determine failure modes? Have you examined the SMART attributes?
     
    izx likes this.
  17. The Lurker

    The Lurker [H]ardForum Junkie

    Messages:
    10,532
    Joined:
    Jul 1, 2001
    MTBF is all BS. I had two of three identical Seagate NAS drives purchased at the same time fail within 4 months of each other. The first 8 months before warranty expires, the second 4 months later. I am waiting for the last one to fail so I can have it replaced too.
     
  18. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,219
    Joined:
    Nov 19, 2008
    With a small sample size you could have 0 failures or all of your drives could fail yet all of the drives of the same exact model in existence may experience a 4% annual failure rate.
     
  19. izx

    izx Limp Gawd

    Messages:
    302
    Joined:
    Jul 30, 2007
    The first few posts in this thread explain what drescherjm aptly summarized. Unless you start off with at least 10,000 identical drives, you should consider the MTBF as BS.
     
  20. jayhopeishere

    jayhopeishere n00b

    Messages:
    10
    Joined:
    Jun 25, 2013
    I don't have any vibration issues. I don't think I have a power supply issue. The power supply never unexpectedly shutdown or reboot unless there is a power outage. I have a 380W earthwatts running 3 HGST 4tb, 3 Seagates 320 GB drives, 120gb Kingston ssd, H700, a i5 cpu, and a EVGA GT 610 graphics card. The graphics card is ok in a power supply 300 and greater. The 3 older seagates hard drives in a software raid are still working.

    The only time the cpu temperature get close to 54 is when I run hand brake. I convert the videos on my old seagates drives and not the hgst when I run hand brake. During this time of the year when I run handbrake my temperature stay below 50.

    I run these drives on a raid 5 array using a perc H700. After a while 2 drives eventually failed in my array. When the first drive failed I put it in another pc and tried to format it. The drive would not format and error out. I also ran the WinDFT tool and don't remember exactly what it said. Currently I am trying to format the 2nd drive that failed. When it get done I will run the SMART tool and let you know what it says.

    I mostly record tv shows to this array with Windows Media Center and HDHomerun. Do anyone think the constantly reading and writing to this array can cause problems with these drives?
     
  21. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,219
    Joined:
    Nov 19, 2008
    I assume you are not doing that 24/7/365.
     
  22. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,219
    Joined:
    Nov 19, 2008
    I would look at the voltages with a digital volt meter.
     
  23. Stereodude

    Stereodude 2[H]4U

    Messages:
    3,285
    Joined:
    Oct 20, 2000
    Not the CPU temperature, the HDD temperatures.
     
  24. jayhopeishere

    jayhopeishere n00b

    Messages:
    10
    Joined:
    Jun 25, 2013
    You are correct. I am not reading/writing 24/7/365. I probably writing about 6 days a week 7.5 - 15 hours/day. I am going to have to buy a volt meter.
     
  25. Stereodude

    Stereodude 2[H]4U

    Messages:
    3,285
    Joined:
    Oct 20, 2000
    :confused: How are you doing that? Are you giving us power on hours or you're continuously recording & playing back / editing from the drives for the number of hours you've listed?
     
  26. izx

    izx Limp Gawd

    Messages:
    302
    Joined:
    Jul 30, 2007
    I don't know which revision of the Earthwatts you have. I had not-so-good experiences with the first Earthwatts EA430 a long time ago.

    Do you overclock your I5?

    With so many sudden drive failures, I'd suspect excessive heat and/or an unstable power supply. The power supply is among the cheapest to switch out and can be reused in a new build, so I recommend -- assuming no overlooking -- you get at least a 430W Corsair PSU (regular on Newegg for $20 AR) and replace the 380W Earthwatts.
     
  27. jayhopeishere

    jayhopeishere n00b

    Messages:
    10
    Joined:
    Jun 25, 2013
    I am giving you hours. I just look at how often I record. Those numbers are not continuously. I record between 7.5 hours to 15 hours a day. I record about 6 hours Monday-Friday on just ESPN alone.

    Also can you tell me how can I measure the temperature of a hard drive behind a array? I have a perc H700.
     
  28. jayhopeishere

    jayhopeishere n00b

    Messages:
    10
    Joined:
    Jun 25, 2013
    I do not overclock.
     
Tags: