On Thu, Dec 4, 2014 at 8:19 AM, Bill Bogstad <bogs...@pobox.com> wrote:
> On Thu, Dec 4, 2014 at 1:06 PM, Edward Ned Harvey (lopser) > <lop...@nedharvey.com> wrote: > >> From: tech-boun...@lists.lopsa.org [mailto:tech-boun...@lists.lopsa.org > ] > >> On Behalf Of Bill Bogstad > >> > >> Not sure if my previous (subject only) note made to the list or not, > >> but the subject line pretty much says it all anyway. Although, it > >> should not matter for answering my question, my motivation is to > >> get some idea of the feasibility of treating disk drives the same > >> way as tape cartridges. i.e. Leaving them on a shelf for extended > >> periods of time and then expecting to retrieve data from them. > >> While there have been public studies in recent years on the lifespan > >> of powered disk drives, I can not find anything other then anecdotal > >> comments about unpowered drives. > > > >[lots of good thoughts on redundant copies, everything decays, etc.] > > > > Unfortunately I don't have performance statistics for hard drive > media... Which is what I think you're actually seeking... > > Yeah, that really is what I want. Without performance statistics, it > is impossible to > know if I would need to keep copies on one or ten unpowered disks. > Should I copy the data to new "media" every year or every decade? > I'm guessing that no one has tried to use disks this way so there > isn't going to be any real world data out there. I suppose even idle > speculation by someone who worked in the drive manufacturing or data > recovery industry would be of interest. For example, I don't have > the physic background to even intelligently guess if individual > magnetic regions/domains on a drive platter will degrade their > neighbors over time, but someone else might. Admittedly, this is only > one part of whether an entire drive will remain readable; but it might > put a useful ceiling on the maximum longevity of an unpowered drive. > > Bill Bogstad > Drives have some other issues in addition to magnetic degradation, such as the possibility of the controller electronics failing and obsolete interfaces (though all mediums have that one). To safegaurd against that, I think, just like anything else, you need to perform routine testing. This would fall into the realm of spinning up the drives, running a full read/write cycle where the contents are read and re-written to the platter, and running checksums against the files. This would help to detect errors as they start to arise. As far as recovering data upon failure, multiple copies are the obvious option, but I would also look into creating Parchive files of the data. You could then store them on separate drives which hopefully would not fail at the same time. Another esoteric option might be to make a RAID 5 or 6 set using a minimal Linux install and software RAID, which you would be able to boot to get all of the volumes back online. I don't have any stats for you, but anecdotally drives often last a few years (3-5, up to 10 with luck). When they fail it's usually a slow process that spreads over the platter, so you would have a chance to make a partial recovery before losing everything. My main fear would be that since the drives are offline, any built-in error correction would not be active so small issues that might have been recoverable may become unrecoverable by the time you spin them up again. ❧ Brian Mathis @orev
_______________________________________________ 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/
