On Backups

September 19, 2008

Backups are good.

Sometimes, hard drives stop working. For example, last week, my hard drive stopped working. (Such a great example!) I didn’t think it would happen, since I have a shiny eight-month-new Macbook Pro. (You hear that, Google? Mac Hard Drive Failure. Mac Hard Drive Failure, Mac Hard Drive Failure.)

Backups are good. (But you knew that already.)

Backups should be easy, so you actually do them. You should test your backups, so you know they work. (But you knew that already.)

My Mac makes these things easy with Time Machine. (It works!)

Backups are one part of a disaster recovery plan. I thought, “Hey, I do backups, I’m safe, I’m a well-behaved geek.” I never thought about the other parts of a disaster recovery plan.

Here’s the problem: It took my friendly local Apple Authorized Service shop six-to-seven business days to replace my dead hard drive.

That’s a week-and-a-half of no access to files, saved passwords, calendar events, and to-dos.

What would you do?

Disaster recovery plans are good.


7 Responses to “On Backups”

  1. haz Says:

    Live in the cloud man.

    “…no access to files, saved passwords, calendar events, and to-dos.”

    I can get all that from any computer, and its only slightly different with the pwds if the computer isn’t one of my own.

  2. Greg Wilson Says:

    Everything on this machine that I don’t want to lose lives in a Subversion repository on a machine with nightly backups. Crude, but effective 🙂

  3. What would I do?

    Subversion at U of T…interesting

  4. Ian Says:

    Three letters… PDA (not that I use one!)

    Although if you don’t back that up regularly, it can be painful too – as a friend of mine learned recently when the screen on his got smashed and it would only flash at him once in while – kind of like a taughnting wink 🙂

  5. George Says:

    “I didn’t think it would happen, since I have a shiny eight-month-new Macbook Pro.”

    Alas, is seems like hard drive failures are well modeled by the Poisson distribution. Since that is a memoryless distribution, the fact that your hard drive was relatively new should not make you any more confident it won’t fail.

  6. aran Says:

    Surely there is -some- wear and tear effect? What is your source for the Poisson claim?

  7. George Says:

    It isn’t really Poisson of course, but it isn’t that far from it. Because at some point there has to be wear and tear and knowing that should let you predict a higher failure rate.

    If you look at the entire lifetime of a drive, it is not really Poisson at all, but if you look in the 2-3 year range it is very well approximated by a Poisson process. I expect if you looked at the 1-2 year range it also wouldn’t be that radically different from Poisson either.

    The point is, the Poisson process is the clear idealization of hard drive failures for reasonable time scales.


Comments are closed.

%d bloggers like this: