One of my system disks died recently, and I was once again saved by SuperDuper, which has saved my bacon twice now.

It takes some time to recover from a backup; this provides ample opportunity to reflect upon one’s backup strategy. While my personal strategy arguably borders on overkill, there was a significant chink in the armor; while I had numerous backups via a variety of techniques, they were all local. In the cases of theft, fire, or earthquake, I was basically trusting to luck.

There are some who advocate periodically copying data to a separate hard drive and placing that drive in a safe-deposit box at a bank. Not a bad strategy, but I’m pretty familiar with my own tolerance for annoying tasks, and that obviously wasn’t a strategy which was going to work for me.

What I really wanted was automatic offsite backup via the Internet, since that wouldn’t involve me hoofing disks back and forth to the bank from time to time. I was already doing some network-based backup for my really critical files to .Mac, but that solution is really only suitable for a relatively small amount of data, and what I really wanted to back up was my Aperture library, which is at present about 45GB worth of digital photos. I’d be seriously upset over losing these pictures, so any online backup service would have to handle them, with reasonable performance at reasonable cost.

There are some very cool services which only handle backup of digital pictures, but I wanted a general backup solution rather than something point-specific.

There are as of this writing three options for general-purpose online backup:

• Mozy
• JungleDisk
• CrashPlan

Each of these products provides some form of trial; I took advantage of these trials, and would encourage others to do the same. In my case, I found CrashPlan to be a perfect fit for my needs.

CrashPlan for the most part runs as a background daemon process. As such, the UI is quite minimalist, since it really doesn’t get used all that often. Here’s the main screen:

Note that backups can have multiple backup destinations, including CrashPlan Central, another computer on the LAN, or a WAN-connected friend’s computer. In my case, I’m using the paid CrashPlan Central service, whereby data is stored in an offsite data center run by CrashPlan, but the other options are free, and allow for ‘buddy backup’ using ‘friends and family’ systems located elsewhere.

Key to consider in use of a network backup facility is the fact that most home cable or DSL connections are asymmetric; that is, the upload speed is typically much slower than the download speed. For example, my DSL connection is rated as ‘up to 3Mb’ download, but only ‘up to 500Kb’ upload. Since sending data to a remote system is all upload, the initial backup is going to take a long time, much as it would take quite some time to suck a bowling ball through a garden hose. In my case, the 52GB I’m backing up took 15 days to accomplish.

That might sound like a lot, but it’s taken me 20 years to accumulate the data being backed up, and the initial backup only runs once, anyway — subsequent incremental backups only send deltas, so it’s all extremely quick once the initial backup has completed.

One very cool feature of CrashPlan is that data is available as soon as it’s been backed up; one doesn’t have to wait for the initial, giant backup run to complete — as soon as an individual file has been sent, it’s available for restore. The restore screen continues the minimalist philosophy, and is quite intuitive:

In order to not monopolize the network connection CrashPlan allows for configuration of network utilization thresholds:

Similarly, CPU utilization can also be configured. However, CPU utilization by CrashPlan on my system is so incredibly low, I doubt that anything but the smallest system would require this option. Identification of changed files and backup of changes is obviously very efficient; it’s basically unnoticeable.

Data is transmitted and stored in an encrypted format, and files are versioned, which allows one to return to a previous version of a file, much in the same way as Time Machine allows for this.

CrashPlan runs on Windows, Linux, and Mac, and CrashPlan running on any OS can back up to CrashPlan running on any other OS. If a destination system is used only as a backup destination, for example, an old PC that isn’t doing anything else anyway, then the software doesn’t even need to be licensed.

CrashPlan is available in a basic version for $25, and a Pro version with more features for $60. This review is based on the Pro version.

CrashPlan Central is as of this writing a little less than $1/GB/year, with a variety of storage sizes avaialble from 50GB to 1TB.

Basically, this thing is set and forget. Brilliantly executed.