My Backup Strategy
A friend recently asked me to help make her backup process more sane. I wrote her a long e-mail, and now I’m rewriting it as a post here for future reference (because I recently forgot what I was doing and had to find that e-mail).
- Backups of 3 classes of assets:
- Irreplaceable and frequently used (a.k.a., FU) — current projects, photos
- Replaceable and FU — stock art, downloaded or ripped media
- Irreplaceable and not FU — archived projects, other external systems (game consoles and such)
- Few total drives
- Always-on incremental backups for IFU data
- Occasional snapshot backups for RFU and INFU data
- Bootable clone of primary desktop
- No NAS
- No cloud
- No DVDs
The last three are controversial, but that’s what we had discussed, so I’m sticking to it for now.
My strategy requires three drives (in addition to the drive in the computer).
Drive 0 — Primary Drive
The one in the computer.
Drive 1 — Incremental Backup of Drive 0
Time Machine backups that are always running for IFU data. Also storage for RFU data and any other disk-based caches you may need, since it’s always connected. I’m also adding INFU here as a simplifying constraint for Drive 3.
You can either partition the drive and tell Time Machine to use one of the partitions, or leave the whole drive as a single partition and just use folders to manage your other storage (which is what I do). Time Machine just needs its Backups.backupdb folder, and it doesn’t care what else is on the drive.
One partition is simpler to manage, but it allows Time Machine to fill the whole drive. You might want to limit the size of the Time Machine backup (do you really need weekly backups from 6 years ago?), which you can only do by limiting the size of the partition it’s on, but you have to guess the right sizes for your partitions. I don’t see tremendous benefit to either scheme. The platter is spinning just as much either way.
It shouldn’t be a problem to leave this drive connected all the time so that Time Machine can run. I’ve been buying SeaGate drives for a while. They’re fast and you can’t hear them over the presumed fan noise from your computer.
Capacity: 2× Drive 0, plus whatever you need for asset storage
Drive 2 — Bootable Clone of Drive 0
This SuperDuper! clone will be the same IFU data as your Time Machine backups (if they both ran at the same moment), except you can boot from it in emergencies, or if you just want to boot your OS on a different piece of hardware for the hell of it.
You can actually run SuperDuper! and Time Machine on the same drive. I considered having two bit-for-bit identical drives that you would leave plugged in all the time that both had Time Machine and SuperDuper! backups, sort of like a crappy RAID 1, so that if one drive fails you have another identical drive. But then one virus that takes out both drives leaves you with no backups.
Frequency: every sometimes
Capacity: equivalent to Drive 0, since it’s a clone
Drive 3 — Non-Bootable Clone of Drive 1
I keep this drive off-site and only back it up about once a month. Here’s my logic:
- Your IFU data exists on Drive 0, Drive 1, and Drive 2. Makes sense, since it’s the most important.
- Your RFU data exists only on Drive 1. But it’s R, so it’s just inconvenient, not catastrophic, if you lose that drive.
- Your INFU data exists only on Drive 1. It’s I, so you don’t want to lose it, but it’s not changing anymore, so occasional snapshots for backup are good enough.
So snapshots of Drive 1 are a good enough backup for your RFU and INFU data, and you would have to lose 3 drives before you needed to get IFU data from here.
Frequency: occasional sometimes
Capacity: equivalent to Drive 1, since it’s a clone
So far, I’ve never had to recover from anything catastrophic, but I’ve used my backup drives for convenience many times. This strategy has worked well for me, and it’s ultimately only cost me a few hundred dollars over the course of about a decade.