Cut and pasted from Wikipedia, not that it has any
Deep history of the subject, but to start the
conversation:
* Data scrubbing?is an error correction technique which
* uses a background task that periodically inspects memory
* for errors, and then corrects the error using?ECC memory
* or another copy of the data. It reduces the likelihood that
* single correctable errors will accumulate; thus, reducing
* the risk of uncorrectable errors.
Some on-topic hardware relation:
I first found out about data scrubbing back when it
was a way to deal with flaky DRAM (what Wikipedia calls
"Memory Scrubbing".) Mid-70's stuff. Later found out
that the Xerox Alto had done it in that era too. Not
sure where it first originated, I would be surprised
if this hadn't been done back to the first storage media.
In the context of large data/disk archives it is reassuring
to have this done not just at the media layer, but above
the transport (e.g. SATA or SCSI or USB bus) layer.
e.g. I store my files as bzip2 format, and every so
often not just make sure the drive thinks it can read
the sectors, but that also it passes a bunzip2 -v checksum
validation. This habit probably dates from flaky SCSI
bus/network protocols in the past that had this odd
way of dropping or mangling random bytes or sectors
without any error indication.
When I'm truly paranoid not only do I count on bzip2 checksums
but also resort to storing MD5's or SHA-1's of the
uncompressed contents.
Every so often I get ultra-paranoid and worry that the file
System is not mangling files but simply losing them. It's been
A long time since I've witnessed this, but storing the MD5's/SHA1's
In a text listing gives me some reassurance. Of course the
Filesystem might not just lose the file but also hide that it lost
The file by removing the file from the listing so I also print out
The listing and attach it to the drive/tape/CD etc... this is getting
To 2001/HAL eras of paranoia to talk about this using a
Computer (never mind locking ourselves into a space pod to discuss
Verbally and then finding out the computer can read lips!)
It is very reassuring to me to do these validations not
Just automatically but manually too. And I even purposefully
Mangle some backup files to make sure my tools would kick
Out a failure.
Probably just paranoia on my part. Wondering if others
might even be more extreme :-).
Tim.