1. 13

  2. 5

    OTOH, wouldn’t an immutable system running from an SSD be essentially immortal? The real drive killer for embedded computers are logs, so a crash handler that will persist logs on reset might be cool. Feels like history is repeating itself, only this time we are building whole interactive workstations instead of temperature regulators and such.

    And consumer electronics are usually not kept that long anyway.

    1. 2

      Most SSDs sold these days have the dial turned way towards cost-efficient capacity over write cycles, reasonably enough. For the industrial/embedded applications mentioned, you can find ones rated for many many more writes.

      My laptop came with a QLC SSD only rated for 200 drive writes. Still, I’ve only used 2% of that in months, despite frequently writing GBs to hibernate it because of a suspend bug. Capacities seem like a factor: harder to fill your SSD 200 times if we’re talking about 1TB not 80GB.

      Generally, surprising how rarely I hear about write endurance actually bricking things. Someone had to swap out 50 consumer SSDs they put in their servers(!), but all around far fewer horror stories than worries.

      Where you really do need higher endurance, to pick a couple random examples Micron seems to claim ~3,000 and ~9,000 drive writes with 3D TLC (huh!) on their 5300 PRO and MAX, and Advantech sells SSDs with high-endurance MLC rated for 30K cycles and SLC rated for 100K. Not claiming these are the best choices, just results of searching around quickly. So you at least have the option of paying for many, many more writes.

      And you might luck out! A German magazine tested a Samsung 850 Pro (rated for 150TB) and it survived 9100 TB of writes. Other drives outlasted their rating too, but nothing quite like that.