Changes between Version 1 and Version 2 of Ticket #939, comment 5


Ignore:
Timestamp:
Nov 14, 2017, 7:54:40 PM (2 years ago)
Author:
Stoat
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #939, comment 5

    v1 v2  
    200200Both are using raw16 as I couldn't see a better way to get the numbers out in readable format whilst suppressing the 0
    201201
    202 I've got a few of these things in 32/64GB and between 3-7 years poweron time. smartctl gives the same results on Linux (rhel and ubuntu). They really are quite crufty devices. :)
    203 
    204 
     202I've got a few of these things in 32/64GB and between 3-7 years poweron time. smartctl gives the same results on Linux (rhel and ubuntu).
     203They really are quite crufty devices with a _very_ low sequential write speed (18/20/40/40MB/s for 8/16/32/64GB) and an endurance of 3000 cycles.
     204
     205advertising sheets quote them at 110MB/s - that's the read speed only.
     206
     207One of our vendors has commented that they tend to simply stop working with little-to-no warning - but none of the vendor appliances have been monitoring bad block parameters - assuming that reallocated sectors was valid. That's why I recommended filtering these returns.
     208
     209What we saw was a dramatic slowdown in write speeds when one of the 64Gb units hit 256 bad blocks - something less than 1MB/s. None of the units has changed the 0x64 "value" field for either bad blocks attribute no matter how many have failed and I'm going to push one through a few write cycles to see if they ever do (I suspect not)