[Noisebridge-discuss] Data integrity, rdiff-backup, Reed-Solomon codes

[dpc]

John Magolske <listmail at b79.net> writes:

>>From which I gleaned...
>
> Current hard drives employ some such error-correction (but how much?
> are some drives better than others in this regard?):
> http://hardware.slashdot.org/comments.pl?sid=634559&no_d2=1&cid=24459631

'most' (where your definition may not meet mine) do a form of
degradation detection rather than error correction. ie, a read will come
back w/ a 'goodness' metric meaning that this sector is going south and
action should be taken. usually the drive firmware deals w/ doing
whatever re-mapping it has to and the metric isn't exposed upwards to
os/driver-land. there is error detection as well, but you're not going
to get raid-like error correction in 'regular' hw that i can rcall (i'm
sure someone will send corrections/flames).

on some optical media (maybe only bd discs) it is even 'scarier' (at
least to my sw person's idea of scary) where the feature being read is
beneatht the laser bandwidth so really it is doing something like
quadrature encoding (someone told me the 'right' term, but i've already
forgotten it).

there will be a large-ish difference between ssd and actual disks. the
failure modes are pretty different (eg, hitting flash read/write cycle
limits vs platter errors vs whatever).

\p
---
No matter how this may shock mankind, the duty of philosophy is to say
everything. - Marquis de Sade