While my wife and I both use Time Machine to back up our Macs, my backup strategy is a little more complicated than that. I have a Mac and a Redhat machine at work which I need to backup, and being paranoid, I keep multiple copies of my backups, just to make sure I don’t lose data.

Dropbox: The majority of my day to day backups run through Dropbox, which I’ve mentioned in the past. I have it installed on the Mac Pro and my black MacBook at home, as well as the linux box and the G5 in my office. I also have Firefox, Mail.app and Safari all set to download attachments into folders in the Dropbox. This way, I have a reasonably synced home folder between all my machines, and most of my ‘hot files’ that I’m actively working on live in 3 places. If one machine dies, there is very little chance of data outage.

Backup Script: I also have a backup script that runs on my two work machines, set to exclude the obvious folders – Dropbox, Music and Photos. It also has file type exemptions for stray Nikon Raw (NEF) files, VM disk images, and the like. The output of file names is dumped to a text file in the backup directory. Once the tarball is ready, the script gzips the tarball and the file list, and then calculates the MD5 hashes of each file. All of these files are moved into Dropbox, which is then synced back to my home machines.

Every month, I copy off the backup files to two sets of DVD-R’s. One gets left at home, and one gets left in the office. This is the über-paranoid backup in case of total system failure of all my computers and Dropbox.

Rsync: I have a periodic rsync that runs to backup the Dropbox to an ZFS backed NFS share. Using the –inplace flag for rsync and ZFS snapshots, only the block differences get saved. It’s a poor man’s de-dupe. IMHO, this should be what Apple does for TimeMachine – not hard linked directories. I’m really looking forward to the ZFS team adding de-dupe into the filesystem directly. I can then expand my backed-up files from the office into an NFS share, and not waste much space.

I also have my music library from my computer backed up into an NFS share that mt-daapd reads from. This was, as I add music to my library, anyone on the network in the house can listen to it – even when my machine is asleep.

Aperture Vault: I also have my Aperture library backing itself up to a sparse disk image I copied over to a CIFS share. It’s much slower than using an directly attached disk, but it works, and with my ZFS fileservers, it’s easy to make volumes bigger. My Aperture library is already 350+GB of data, so I need the flexibility that ZFS offers