i have no experience with hard drives being stored unpowered for a long time
and then re-energised.
but i do have significant experience with tape.
in particular, i have done exactly what jack talked about.
that is, i wrote a few thousand tapes carefully
(that is, i wrote them with application-level checksums and then
some days later, read them back verifying checksums),
stored them in a machine room, and then several years
(that i can’t remember exactly but 6ish, i think) read them back
(so i could copy them to more dense tape).
from memory, i had trouble with exactly one tape, and
i had a read error on two different 256KB blocks on that tape.
this is out of maybe 2100 tapes (these were Sony 50GB AIT tapes).
i have written close to 2000 LTO tapes as well (LTO-1, -2, -4),
and had three tapes develop read errors, although one was
caught during the read-verify part of teh recording process
(and was due to a faulty media batch).
to put it in perspective, i have had 5x as many file corruptions
in linux disk filesystems as i have from tapes, even though i had
maybe 100x as many bytes on tape.
if you care about your old data, i still know of no better solution
than to keep it on tape, and every 5-8 years or so, copy it to new
media. this has, for a long time, been a nearly constant effort
as the increase in media density outstripped the increase in data.
and each copying effort made us reassess if the data was worth keeping.
of course, this has been blown up because of “big data”
(throw another petabyte on the barbie!). the other worrying trend is
that while media/disk sizes are increasing rapidly, the interface/drive
speeds are pretty much static, which means that even if you kept
all of last year’s data, you don’t have enough time to read it all.
sorry for the polemic!
andrew
On Dec 5, 2014, at 12:58 PM, Doug Hughes <d...@will.to> wrote:
>
>
> On Fri, Dec 5, 2014 at 3:22 PM, Jack Coats <j...@coats.org> wrote:
>
>
> Archival grade DVDs seem to have a 50 year 'sure life'. Magnetic media
> (thinking tapes, 5 years, and reel tapes need to be 'exercised' annually to
> extend their lives). 'Flash' media like memory sticks need to be 'refreshed'
> by plugging in, even if data is not accessed, every year or two before bit
> rot starts.
>
>
>
> All these are 'rules of thumb', and detailed or specific cases can and will
> be quoted all over the place. When I worked for 'big oil' they had 50 years
> of mag tapes (from the oil field seismic data) that was stored well
> (exercised, in proper thermal environment, out of the weather, inside dust
> cases, etc), but they found a 50% failure rate on data 5 years old and older.
> They started just trashing their oldest seismic data. Re-digitizing what
> they could (in some cases paper maps and logs were better than digital data).
> This also made the data they had more useful. But even that had its
> failings. Old data was not of high enough resolution, or not measured in the
> spectra near what is currently used. So much old seismic data was kept, but
> new was acquired (at high cost in many cases) if possible to allow it to be
> useful in current exploration / re-exploration of potential fields.
>
> I'd buy into disk inactive life of ~5 years, but LTO, Ultrium/IBM, and
> StorageTek have a published archival life of 30. I wouldn't count on that,
> per se, but certainly better than 5. The current technology of the tape
> substrate is quite amazing stuff. (think Kevlar with chemically welded Barium
> Ferrite magnetics). Yes, you still want to run periodic data verifies, but
> that is trivial with tape. IBM and Oracle/StorageTek are arguably better
> technologies than LTO in terms of durability, bit error rate, and density.
> LTO pretends to be catching up but it's own projections are falling pretty
> far behind unless something miraculous happens (don't count it out. We're
> hopeful, but realistically the other is still better)
>
> It sounds like they had a bad batch. What technology? It has come a long way
> even in the last couple of years. Plenty of companies recover from and still
> rely on tape, despite there being a few with a bad experience story. The
> story itself should be able to stand under scrutiny.
>
> _______________________________________________
> Tech mailing list
> Tech@lists.lopsa.org
> https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech
> This list provided by the League of Professional System Administrators
> http://lopsa.org/
-----------------------
Andrew Hume
949-707-1964 (VO)
and...@research.att.com
_______________________________________________
Tech mailing list
Tech@lists.lopsa.org
https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
http://lopsa.org/
