I did a bunch of reading/research/experimentation with online backup services, yesterday.

[http://www.carbonite.com/ Carbonite] is a good online backup service, but the interface is very basic. They advertise “unlimited” data storage for $5 per month. (Although they have a so-called “god clause” in their Terms Of Service (i.e., “we can terminate your account for any reason at any time”), and presumably storing “too much” data could be perceived as abuse, there’s no information on what “too much” data is. Users in comments report storing 50 GB in the service with no issue, and one fellow had uploaded 500 GB (over a long period of time). Carbonite’s simplified interface gives the impression that it “just works”, and in fact reports are that it’s reliable. It allows you to choose files/folders to backup by right-clicking on them, and indicates backup status of each file with an icon overlay, a la [http://tortoisesvn.net/ Tortoise SVN].

As an alternative online backup solution, there’s [http://aws.amazon.com/s3 Amazon's S3 service], which comments speculate Carbonite is likely using as their back end. Rates are 15 cents per gigabyte per month for storage, and 20 cents per gigabyte transferred. S3 is just a raw Web Service (no program, no GUI, nothing), and clients for it are not yet mature. (But in a way, that’s part of the fun of it.) One way to use it as a backup system is to use two free programs, [http://www.jungledisk.com/ Jungle Disk] and [http://www.acs.uwosh.edu/novell/netdrive.htm NetDrive], to get a mapped drive letter for your S3 storage account (under Windows). Then use a backup or sync program to move data to the mapped drive. I gave [http://www.2brightsparks.com/syncback/ SyncBack] a try, but the problem is that since SyncBack stores no data about your backup history, it needs to do a full enumeration of all the files in both the source and destination every time it makes a backup (to determine which files to back up). That’s no problem for sync’ing from one hard drive to another over a LAN or USB or some other fast connection, but with S3 determining information about a batch of files is inefficient. In order to accurately determine if something actually changed, SyncBack currently is downloading the entire file. A better solution (I have yet to try) might be [http://allwaysync.com/ AllWaySync], which stores information in an XML file about past sync operations. That way, the files themselves don’t need to be examined, just the XML file(s). This should be fast. Although Carbonite is easy, it only backs up data files, not programs by default. Still, that’s good data insurance in case of fire or vandalism or an outdated image file, but keeping a Ghosted disk image around is probably better for the most typical cases of data loss which would be (1) hard drive failure or (2) bad virii. In essence, then, the compexity of getting an automated S3 routine working may not be any better than imaging your disks every once in a while, combined with if-all-else-fails “data insurance” through Carbonite or a simple service like it.