The Mean Time

In the hardware world, there is a wonderful phrase: mean time between failures (MTBF). It means, roughly speaking, the average amount of time one should expect before a hardware failure. It’s an average, of course, so your piece of hardware that has an MTBF of 3 years might fail after just 2 months. Or it might just keep on humming along for 9 years. (The hard drive on an early Tivo we got in 2000 is still going strong, for example.) Until just recently I had not had a serious hard disk failure in at least 5 years. I can’t actually remember the last one, although the clicking sound my drive was making was not unfamiliar (disturbing, but not unfamiliar), so I know it has happened.

Yeah. The hard drive on my iMac failed a week ago. This had the potential to be quite devastating. It wasn’t, but not because I’m doing so well at my backup strategy. Even though I know damn well what I should be doing – [embarrassing geek confession] I have not quite been on top of it. Really, “Cosmic Whole-House Backup Strategy” has been on my big ToDo list for, umm, years! This particular failure wasn’t devastating mostly through dumb luck.

There is one irreplaceable set of data on that drive: the photo database. It’s not even all of my photos, only from 2007 or so, but it’s over 8000 of them, mostly of TLG. That I have been backing up manually to an external drive every week. It’s a hard calendar item for me. (GTD fans will know what that means.) So, at most I lost a week’s worth of photos. I have restored that database onto my Macbook just to prove that I can, and everything seems intact. The other set of data that’s particularly crucial is my Omnifocus database. That is replicated to both my iPhone and my Macbook. I haven’t gotten that back onto the iMac and up and running again, so I’m not 100% comfortable about that yet. But even so, manually recreating it – when I can look at a copy on another machine – would not be the worst thing, just a bit tedious.

And, I had a bootable image of that drive from April. So I lost anything that’s not photos and that’s not Omnifocus between April and a week ago. But, you know, I’m not actually doing a lot of creative work with my iMac lately because I have no time. I live in the browser and in Aperture mostly, so I didn’t really lose all that much. Sheer dumb luck.

The day after it failed I decided to break out of ye olde analysis paralysis and just ordered a Time Capsule from Apple (refurb’d, 1 Terabyte). The Macbook is already on a regular backup schedule with it. We got the iMac back last night (with a brand new empty drive), and I’m doing the first Time Machine backup of that (figure it’ll never be smaller than it is right now) as I type this. That’ll solve one problem. TheGuy ordered a couple new big external drives that we’ll use to manually backup/create bootable images of the Macs on to. Probably that will be a monthly hard calendar item. That only leaves my PC, which I’ve been aiming to retire soon, anyway, in favor of a virtual machine on the iMac. Related to that, though, I’m not having Time Machine backup the virtual machines on my Macbook or iMac – I may just copy those over manually when we do the bootables and call it good. I don’t store much data on those things – they’re primarily used for work. I am reading that the MTBF of the Time Capsules may not be that great, which is a concern. But what we’ve got now is better than what we had a week ago, so that’s progress!

Once that’s all in place, the last remaining challenge will be offsite backups. If the house burns down, where’s all my data? I think I’ll try to use JungleDisk or something similar for that. But the question is whether I want to be more parsimonious about what’s backed up into the cloud – for bandwidth reasons. If anyone’s got any great offsite backup strategies that don’t involve shuttling hard drives back and forth between office and home, please share!