Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Tuesday June 17 2014, @06:36AM   Printer-friendly [Skip to comment(s)]
from the write-right-rite dept.

In 2013 an experiment to test the endurance of SSD consumer drives by writing 6 PBytes onto them was launched. They are now one petabyte into the experiment. The performance is the same. The casualties slowed down a little toward the very end, and they ceased to function in different ways. The Intel 335 Series and Kingston HyperX 3K provided plenty of warning of their imminent demise, though both still ended up completely unresponsive at the very end. The Samsung 840 Series, which uses more fragile TLC NAND, perished unexpectedly. It also suffered a rash of cell failures and multiple bouts of uncorrectable errors during its life. The sample size is too small to draw any definitive conclusions. All of the six SSDs exceeded their rated lifespans by hundreds of terabytes. The fact that all of them wrote over 700 TByte is a testament to the endurance of modern SSDs.

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 0) by Anonymous Coward on Tuesday June 17 2014, @07:03AM

    by Anonymous Coward on Tuesday June 17 2014, @07:03AM (#56262)

    Kadir beneath Mo Moteh! Headline. Shaka, when the walls fell.

    • (Score: 0) by Anonymous Coward on Tuesday June 17 2014, @07:15AM

      by Anonymous Coward on Tuesday June 17 2014, @07:15AM (#56265)

      30, 56, 99 are correct. Limited 4 and 8 are missing.

  • (Score: 4, Interesting) by mojo chan on Tuesday June 17 2014, @07:36AM

    by mojo chan (266) on Tuesday June 17 2014, @07:36AM (#56267)

    From TFA: "The stream of sequential writes in our endurance test isn't indicative of real-world client workloads."

    So what use are these figures, except for the very few number of people who do masses of sequential writes? Random writes are what really kill SSDs because of write amplification.

    --
    const int one = 65536; (Silvermoon, Texture.cs)
    • (Score: 5, Informative) by evilviper on Tuesday June 17 2014, @08:26AM

      by evilviper (1760) on Tuesday June 17 2014, @08:26AM (#56273) Homepage Journal

      From TFA: "The stream of sequential writes in our endurance test isn't indicative of real-world client workloads."

      So what use are these figures, except for the very few number of people who do masses of sequential writes?

      Just because their test isn't a simulated client workload, doesn't mean they are completely uncorrelated. Plenty of tests are done the same way... Extremes are used to allow testing in reasonable time-frames, then the results are extrapolated down to the equivalent levels of average usage, to give a real-world time-frame on everything.

      --
      Hydrogen cyanide is a delicious and necessary part of the human diet.
  • (Score: 1) by NowhereMan on Tuesday June 17 2014, @05:23PM

    by NowhereMan (3980) on Tuesday June 17 2014, @05:23PM (#56539)

    So, how does this compare to current disk drives in terms of life expectancy?

    • (Score: 3, Interesting) by broken on Tuesday June 17 2014, @06:51PM

      by broken (4018) on Tuesday June 17 2014, @06:51PM (#56581) Journal

      Because of their different failure modes, I don't think you can directly compare them to standard disk drives. SSDs can wear out from excessive writes. Standard drives have no real write limit. Instead they tend to wear out from start-stop cycles and mechanical wear on the moving parts, in addition to environmental factors including temperature, vibration, and shock (e.g. dropping a laptop). If you know that you are unlikely to hit the write limit of an SSD, my understanding is that it will probably outlast a traditional hard drive.

      I'm not a hard drive expert, so to anyone who is more familiar with these issues, please chime in and correct any misapprehensions I may have.

      • (Score: 1) by NowhereMan on Wednesday June 18 2014, @03:36PM

        by NowhereMan (3980) on Wednesday June 18 2014, @03:36PM (#56989)

        Okay, fair enough, but I don't have a context to understand the level of abuse these SSDs are going through so looking at it from another way, what kind of abuse would I have to put a hard drive through to equal what these SSDs are going through?

        • (Score: 2) by broken on Wednesday June 18 2014, @07:58PM

          by broken (4018) on Wednesday June 18 2014, @07:58PM (#57110) Journal
          The intent of this test is not to "abuse" the SSDs, but to measure their write limits. The number of write cycles is often the limiting factor that determines the longevity of an SSD in the real world. For standard hard disk drives, the age of the drive is one of the limiting factors - drives used half as much don't usually last twice as long. But if I wanted to see how much I could shorten its life by extreme usage, I would try some combination of repeated power cycling of the drive (giving it time to come up to operating temperature and then cool down), and doing repeated seeks between the inner and outer tracks.
  • (Score: 3, Interesting) by tonyPick on Wednesday June 18 2014, @07:09AM

    by tonyPick (1237) on Wednesday June 18 2014, @07:09AM (#56812) Homepage Journal

    The really fascinating thing here (for me) is the failure modes - the fact that these drives went into brick modes is more than a little scary; one of the advantages I'd hope a quality SSD would give you is a more graceful degradation than these tests show. Drives that fail without warning, or deliberately fail in non-readable seems bad (although no worse than spinny-platter disks to be fair).

    This bit from TFA bothered me:

    " When the power is cycled in this state, a sort of self-destruct mechanism is triggered, rendering the drive unresponsive. Intel really doesn't want its client SSDs to be used after the flash has exceeded its lifetime spec. The firm's enterprise drives are designed to remain in logical disable mode after the MWI bottoms out, regardless of whether the power is cycled."

    Wait, what?! The consumer versions *deliberately* brick themselves, even with valid readable data!?