While writing my article on backups (and it’s followup) I decided some of my practices weren’t what I wanted them to be. The primary issue was the online catastrophic backups, which used Jungle Disk as a front end to Amazon S3 for storage. I really like the setup — Jungle disk was almsot flawless in doing what I needed the way I wanted it done, and S3 was reliable and backed by Amazon, so I didn’t need to worry about the “not here tomorrow” problem you sometimes have with startups.
But there were a few negatives: the cost — I was spending about $17/mo on storage costs with S3, plus about $1.50 a month to Jungle disk for their advanced features. I also worried about the occasional delays (I was only uploading about a gigabyte of fresh data a day, so a few days of heavy photo shooting could cause backlogs before the data was stored offsite; uploading the network is an option, but costs. Everything costs…) and finally the time it would take to recover via online recovery if I ever needed the backed up data bothered me.
So I decided to move to a simpler strategy: clone my disks and keep them offsite. Over the weekend I disabled and deleted the S3 store and turned off my Jungle Disk setup, and ordered a new drive.
Not an offsite drive, which might surprise you, but a new internal drive for the Macbook, a 340 Gig Hitachi 7200RPM drive. Why, might you ask?
Well, let me tell you: if you read the previous pieces, you saw where I noted that one of the best ways to never NEED your backups (always my preferred policy!) is to replace my primary disks on a regular basis; my laptop drive was (over)due, so it made sense to replace it before it failed. By upgrading to a larger internal disk, I could take the files I’m currently storing on an external firewire drive and put them back on the laptop drive.
That simplifies my computing universe — fewer spinning things hooked up to computers, fewer places to misplace data and one less “i have that data, but it’s not with me” opportunity. And it means I can repurpose that firewire drive for my offsite backup.
Amazon had a nice deal (about $75) on a Western Digital 320gig 7200RPM drive with a 16 meg cache (hmm. old phart warning: the first hard drive I ever bought — for a Mac 512K that plugged into the floppy port!) was a ten Megabyte drive that was wonderful and had more disk than I could ever think of using, especially compared to floppies… Now that drive is smaller than the performance cache on a hard drive… wow).
I could have gone as far as 500 gig, but that changes the price/performance,a nd even with copying all of the files off my secondary drive, I still have 150 gig free. By the time I start worrying about the disk filling up, the bigger drives will get cheaper, or I can simply buy a nice cheap external and split it up again. But for now, I’m happy on a single drive, a single backup drive, and a single offline external archive drive. Plus backups of each, of course.
I wired up the raw drive via USB to my laptop and used superduper to clone my primary drive. Then I let everything sit for two days with the new drive spun up, refreshing it daily with superduper — because infant mortality on your laptop drive really sucks, and giving it a couple of days before opening up the guts and swapping things is a bit of insurance. Just saying — nothing like putting your laptop back together and having it fail (or not boot).
A bonus feature of this change: the old drive in the laptop was a 5400 RPM drive. Upgrading to 7200 RPM improves the I/O characteristics and speeds the overall performance, especially if you’re doing things that eat lots of virtual memory (like, oh, photoshop or lightroom, or running both). On a mac, check /var/vm and see how many swapfiles you have and how big they are. The larger your VM set, the more your disk is going to affect overall performance, and in many cases, a slow disk is the real problem to performance, not lack of RAM. hint: people how are proud of NEVER REBOOTING THEIR COMPUTERS are never re-initializing their VM environment. Silly boys. I’ve seen major complaints about slow performance in firefox and photoshop magically disappear on computers that were simply rebooted. Just saying.
So for $75, I remove a $20/mon charge to pay for the online backups, I add enough disk to my primary computer to keep me for at least six months or longer, added a faster disk to speed up overall performance of the computer, and I really didn’t do anything to weaken my backup strategy. All I need to do is remember to bring the catastrophe disk home once a month, refresh it, and get it back offsite again; I can choose to do it more often if I finish a significant project, too. And that’s very manageable for me — just put repeating tasks in your calendar to nag you, right?
There is one other upgrade I’ll need to make, which is my bus-powered backup drive is now too small to fully backup my primary drive, but I don’t need to worry about that for another 50 gigs of data or so; that’ll cost me another $75 (or less by then) down the road.
So I now have my primary disk (320 gigs) with a time machine backup and THREE superduper cloned backups (one offsite, one nightly, one weekly); I also have a pair of 500 gig drives which store all of my long-term offline archival stored data; they’re clones of each other, and one is stored offsite. These hold the files I don’t ever expect to touch again but won’t throw out for a while just in case (the import backups of all of my photo shoots, for instance), so they are in fact more redundancy. Just in case, you know?
The old laptop drive? It goes into the anti-static bag and gets filed with all of my yearly paperwork and taxes; it’ll stay there until I dispose of this year’s paperwork down the road, at which point the drive will have a bad date with a big hammer and some other power tools and get tossed in the trash. I much prefer my hard drives not end up in hands of strangers (if I do give a drive to somene else, it’s generally after a multi-pass, multi-pattern write/erase sequence, but drives are cheap enough, I kinda feel they’re not worth losing track of until you know they’re destroyed, and there’s no sense destroying until later, because you never know when you might need it…. It hurts nothing to wait…)
So I’m following my own advice and fixing my own backups — again. Using simple technology to reduce “failure by complexity”, using as few mechanisms as reasonable, using multiple redundant backups and backing up on different time sequences to avoid the problem of corruption not found until too late — and keeping copies off site, without being anal about it.
It’s probably overkill, but disk is cheap enough now that it’s cheap insurance. And I’d rather have one too many backups than one too few.