Is it safe to store data on a hard drive for a long time?

Is it safe to backup data to a hard drive and then leave it for a number of years?

Assuming the file system format can still be read, is this a safe thing to do. Or is it better to continually rewrite the data (every 6 months or so) to make sure it remains valid?

Or is this a stupid question?

Solution 1:

I wouldn't trust important backups to any single device for any significant length of time.

I've had plenty of CDs that couldn't be read after a while. (Cheap ones, admittedly, but I'm leary of the longevity claims made.)

I've had hard disks silently corrupt data.

I seem to remember I've even had SSD failures, although with a low number of writes I'd expect them to be pretty reliable.

Aside from all of these things, using a single copy means you've got no protection against physical disasters: fire etc. If you have multiple copies, you can separate them physically. Ideally I'd take some number (e.g. 3) of copies and run a checksum (I usually use MD5) periodically over everything. If one of the copies becomes corrupt in some way, if you've got multiple other copies you should be able to trust the majority, and create a new backup to replace the corrupted one. (Of course, if you keep the correct checksums in a separate place, you could trust even a single backup which still gives the right checksums, as the canonical source for replacements.)

Of course, how much trouble you go to depends on the value of the data. My personal home data is only backed up on a RAIDed NAS. My work data is in Google datacenters, which I trust fairly strongly :)

Solution 2:

Given your other options for backup, HDD is the safest way to go. Other options include Magnetic Tape, SSD and optical media.

Let's examine the pitfalls of each:

MT: More prone to erasure when exposed to a magnetic field than a HDD. Readers are also becoming harder and harder to find. You don't want to come back in 5 years and find that there's no way to remove the data from your medium.

SSD: Reliable in that there are no moving parts. They are prone to electrical degradation after several read/write cycles which is troublesome and potentially dangerous. The likelihood of losing data while the drive is not in use is slim, however.

Optical Media: The least reliable of the bunch. They're extremely prone to physical degradation (bending/warping) and it requires very little to throw them out of their deflection spec. Further, the encoding scheme used to write data to most optical media is rather complex, creating a greater likelihood of single element failure leading to unreadability.

HDDs: Solid, sealed devices. Can be damaged by physical shock more easily than most of the above devices. Has precise mechanical parts that can lead to failed read/writes if damaged.

The benefit of HDDs, however, is that they ARE sealed. All of the moving parts are stored in an air-filtered enclosure. The magnetic stability of the bits on the disk is quite high and unlikely to change.

Further, if the mechanical parts fail, it is possible to have the platters removed and the data recovered from them directly.

There's no perfect option, but of the imperfect ones, HDD would probably be your best bet.

Solution 3:

I would say you should recycle the media every other year or so - that is, replacing the drive, disc or tape with whatever there is to replace it with and keep more than one copy.

Few things lasts forever, optical media can degrade rapidly depending on quality, method of writing to it and environment where it's stored. Mechanical parts can always fail or there could be bugs in the firmware that is related to time or to wear and tear.

I've pondered over your question often, it would be convenient with something that is guaranteed to stay working for say 5 years. There's tapes and other form sof backup media rated for 10 or more years but I'd never trust that, at least not without a decent amount of redundancy (several copies on different batches).

Keeping the data fresh and continually recycled seems to be the reliable way to go - that way you get to test it regularly as well.