If I remember correctly it was during 2007 that Google released a paper on the reliability of the hard-drives that they use in their data centres. Google, being Google, quite obviously have a lot of hard-drives working away 24/7 in their servers and data-storage centres, and therefore didn’t need to be interactive with the public to carry out this analysis. Their purpose for doing this was to discover trends and traits regarding performance and failure rates in these components.

They found that, on average, one disc out of 14 that was 2 years old or more, failed. I don’t imagine that Google would use cheap and nasty hard-drives either; so we can assume that the discs studied were of at least average quality.

Numerically we can work both forwards and backwards from that statement with at least some degree of accuracy and say that for discs that are over 1 year old there would be 1 failure in 28, and for discs over 3 years old there would be 1 failure in 7. Working this way, the figures indicate that, whilst the usual maximum life of any hard-drive in use 24/7 is six years, there is the odd 1 in 333,333 case where the drive could continue to function perfectly for 25 years or more.

(As partial testament to this I have a few working hard-drives in my spares cupboard that have less then a megabyte’s capacity and have been removed from people’s computers in the last 2 years while upgrading the hard-drive to a secondhand PATA drive of around 20MB capacity bought off eBay for a few pounds sterling. Those replaced drives must be getting on for at least 10 years old!)

(Note: I build computers only: I normally don’t do any upgrades or major repairs of any kind for anybody (Because all I get is hassle that the computer’s not exactly as it was before and/or that the person doesn’t like this or that new or different behaviour or feature… (Even if the machine’s better than it’s ever been before.) Moan, whinge…), except either under warranty and/or for close friends and/or long-time customers that I trust. – Even then they must assure me in writing that they have a full backup of all their data at home beforehand, just in case the worst happens. (A lost or defunct rebuild can be put right; even though it may leave me out of pocket: A hard-drive that goes down with all of a person’s data on it that is not backed-up can never be fully replaced.))

- Which brings me on to my usual nag: Always always back up your data regularly or you’ll end up very sorry that you didn’t. It’s a case of when not if. New hard drives are fairly cheap and getting cheaper by the day, whereas a large proportion of your data can only increase in value over time.

Back to the main thrust of this article: The Google study also indicated that the SMART technology, (Self Monitoring, Analysis and Reporting Technology.) wasn’t exactly reliable in predicting when or if any given drive was about to fail. They found that a number of drives reporting SMART warnings lasted for quite some time, even years, while other drives gave no SMART warning prior to failure.

Those results indicate that you can never tell exactly when your hard-drive is going to fail. (I’ve even heard of a brand new hard-drive failing on a brand new computer (Not on one of my builds. ) within the first day before now.)

What should people do, then, to avoid the worst happening?

Answer = Hard-drive failures are unavoidable, and can happen at any point, either with or without warning.

If that doesn’t scare you then you’re made of stronger stuff than me; and that’s saying something. – So there is but one answer: <nag time> Backup, backup, backup. </nag time> – End of story.