CrashPlan: Backups Revisited
In the course of writing two editions of “Take Control of Mac OS X Backups” and several magazine articles on the same subject, I’ve tried a lot of Mac backup applications. Although the details of their features and implementations vary greatly, I haven’t seen anything truly revolutionary for a long time. Mac OS X 10.5 Leopard’s much-discussed Time Machine feature will certainly apply a snazzy interface to file restoration, but its underlying mechanism of archiving files to a hard disk or server volume once a day is pretty standard.
So when I heard about CrashPlan from Code 42 Software, released last month at Macworld Expo, I was expecting more of the same: just another tool to add to the dozens already in my lists and feature comparison charts. I was pleasantly surprised to discover that CrashPlan is much more than that. It does some cool things I’ve never seen in a Mac backup program before, and it does them in a wonderfully easy-to-use fashion.
Most backup software follows a similar pattern: you tell the application which files, folders, or volumes you want to back up, where you want the backups to go, and what options you want to use (such as compression or filtering by file name). Then you either click a button to perform an immediate backup or schedule recurring backups. In CrashPlan, the file choices and options are almost an afterthought; it’s the destination you’re asked for first, and that’s where the program begins to distinguish itself.
Smarter Online Backups — One option is to back up your computer over the Internet to CrashPlan’s servers, called CrashPlan Central. The main benefit in doing so is that your data is stored safely off-site. It’ll still be there if your house burns down, and you can even access it remotely from another computer if necessary.
In general, despite this advantage, I’ve found Internet backup services underwhelming, because they’re slow (even with a fast broadband connection) and expensive, and because restoring files is problematic if your Internet connection disappears. However, CrashPlan’s pricing is vastly lower than its competitors: you can store up to 50 GB for $5 per month, and additional space is only $0.10 per GB. In addition, you’re charged only for the amount stored at the end of the billing month, not the total amount of data transferred.
Like other online backup programs, CrashPlan compresses and encrypts your data before sending it over the Internet. And, like the others, it’s constrained by the laws of physics; compression notwithstanding, online backups can in some cases take days or even weeks to complete. But CrashPlan does three unique and interesting things.
First, it gives precedence to files that have changed recently. So, if you’re running your first full backup and it takes three days to finish, but in the meantime you make major changes to a document you’re writing, CrashPlan almost immediately backs up the new document rather than waiting until the rest of the queue has been processed.
Second, it can store multiple versions of files that have changed, and can optionally do so right after a change takes place, rather than having to scan every file on your disk before each run; you can set the number of minutes that elapse between when the file is saved and when CrashPlan backs it up. This means that, from the moment you begin using CrashPlan, it can archive multiple versions of your most important and actively used files – somewhat like a version-control system – without waiting to run on a fixed schedule.
And third, if a file changes, CrashPlan doesn’t make a new copy of the entire file, but copies only the bytes that have actually changed. This means that incremental backups run very quickly and that storage space is minimized – two attributes that are always welcome, but especially so with online backups. As a consequence, if you were to back up, say, the 10 GB disk image you use for Windows XP in Parallels Desktop, you could update it every day without worrying that you’ll overwhelm your backup media. In fact, CrashPlan goes a big step further: if more than one computer in your home or office contains a file identical to one you’ve backed up already, CrashPlan stores just a pointer to the original backup file, not another
copy. So you can effectively store much more data in a much smaller space than with other backup solutions.
The potential downside to this partial-file approach, however, is that in order to reconstruct a given version of a file, CrashPlan needs access to all its constituent pieces. Although the program performs extensive data verification to make sure nothing is lost, conceivably a disk error or other snafu could damage some portion of one of your backed-up files, making it impossible to reconstruct the whole thing later; this is much less of a worry when your backup program stores complete copies of each file.
Peer-to-Peer Backups — CrashPlan’s servers aren’t the only destination you can choose for your backups. In fact, the company’s FAQ talks candidly about the fact that CrashPlan Central is neither the fastest nor the cheapest choice. A second option is a friend’s computer, which could be almost any other computer in the world with an Internet connection. CrashPlan has clients for Mac OS X, Windows, and Linux, and files from one operating system can be stored on another. All your friend has to do is install the software (it’s free to use if you’re only storing someone else’s backups) and you can then add that friend’s hard drive as a backup destination for
your files. Your friend can specify how much disk space is available for you to use, and all your data is encrypted in such a way that your friend can’t read your data. Neither you nor your friend requires a static IP address to pull this off, though in some cases you may have to fiddle with a firewall or router to let the data flow correctly.
What you can do over the Internet with a friend’s computer, you can also do on your local network with another computer of your own – or more than one. And once you’ve set up computer A to be a destination for computer B, you can also back up from B to A with a single click, making the backups mutual. This means that whether you have two computers or dozens, each can back up to the others, quickly and automatically (assuming you have enough disk space).
Most backup software can be persuaded to back up to another machine on the local network by using file sharing to mount one computer’s disk on the other. More-advanced software, such as Retrospect, bypasses all that hassle and also provides greater control by using a client-server system. But CrashPlan is the only cross-platform backup software I know of that provides true peer-to-peer backups, locally or remotely, out of the box – and that it does so without requiring explicit scheduling is a huge bonus.
Bells, Whistles, and Red Flags — Beyond the basics, CrashPlan offers most of the features I consider important in any archiving backup program. You can choose whether to have the software run all the time or only during certain times; select how many versions of each file you want to keep and when, if ever, to remove files deleted on the source disk; select only certain files or folders or exclude files with particular extensions; choose files to restore by date, by filesystem location, or using a search; restore backed-up files to their original location or another spot; and either rename or overwrite existing files during restoration. You can also throttle the bandwidth used for online backups, and can
choose different thresholds for when you’re using the computer and when it’s idle.
Nevertheless, for all this backup goodness, CrashPlan has several flaws and some curious omissions. For one thing, bafflingly, it lacks the capability to back up directly to an external hard drive or even a network-attached storage device; at the moment, the only valid destinations are other computers (and only the users of those computers can specify on which local storage device backups are stored). So if you have just one computer and can’t – or don’t want to – back up over the Internet, you’re out of luck; if you do back up over the Internet, you’ll be stuck with slow restore speeds. Code 42 says that remedying this is at the top of their priority list.
Another issue is that CrashPlan, like Time Machine and Apple’s Backup 3.1, was designed to do archives well but not to create bootable duplicates. In my opinion, archives and duplicates are the two pillars of a complete backup system, and having just one puts you at risk. For example, even if you have excellent archives, a hard disk crash could leave you unable to work for hours or days while you repair or replace your drive and restore all the necessary files; duplicates let you get back to work immediately. In addition, bootable duplicates can save your bacon in the result of a system upgrade gone wrong. So you’ll want another tool, such as SuperDuper, to cover that aspect of your
CrashPlan is written in Java, and as is often true with Java applications, it leaves something to be desired in the look-and-feel department. That is to say, it’s reasonably pretty, but looks more like a Web page than a proper Mac application. There are no menus or keyboard shortcuts, and some of the controls are ambiguous. Although backups run in the background even when the CrashPlan application isn’t open, you get no feedback that anything is happening, except possibly your Internet connection slowing down (depending on your settings). Code 42 says they’re working on a solution to that.
On a few occasions, I noticed that my Mac’s CPU got stuck at 100 percent for long stretches of time even when I didn’t have any other applications running and CrashPlan wasn’t actively backing up; Activity Monitor informed me that Java was monopolizing the processor. According to Code 42, a few other users have had similar problems, and they’re looking into it.
These, and some other minor bugs, are about what one would expect from the first commercial release of a new application, and I fully expect to see improvements in the near future. Among the other things I’ve requested are greater flexibility in excluding files from a backup (say, by size or date modified) and a grammatical overhaul: the program uses “backup” as a verb when it should be “back up” (as in “Files to Backup” [sic]).
Conclusions — Despite these problems, CrashPlan is the first program I’ve seen that has made me want to reconsider my own backup strategy, because it offers possibilities that previously did not exist or were awkward to pull off. I’m not quite ready to give up Retrospect for backing up the Macs in our home, but I can easily imagine that happening if and when a few of CrashPlan’s outstanding issues are ironed out.
CrashPlan comes in two editions. CrashPlan Pro (described here) costs $60. The non-Pro version of CrashPlan, which costs $20, stores just one version of each file and doesn’t look for file changes in real time. The Pro version also includes free updates for a year and priority tech support. A free 14-day trial is available.
Staff Roundtable — [Adam Engst] I haven’t had a chance to look at CrashPlan myself yet, but I’m extremely intrigued by Joe’s description of its peer-to-peer backup approach. I suggested that Apple should do something along those lines in talks I’ve given and in “Apple Computer, Going Forward into 2004” (2004-01-05), where I wrote:
“I’d like to see Apple build peer-to-peer file sharing technologies into Mac OS X so a network of Macs and Macintosh users could become more powerful and flexible than the mere sum of its parts. Multiple copies of shared files could be stored across multiple machines, eliminating the problem of a turned-off Mac or traveling PowerBook. Browsing for servers would become a thing of the past; shared files would simply be accessible at all times in what seemed like a local set of folders. Such a system could even automatically maintain backups of data that wasn’t shared by distributing it in encrypted chunks around other Macs on the network. All this would work best over a high-speed local network, but designed and implemented
properly, nothing would stop it from extending to the Internet.”
When I get to testing CrashPlan, I’ll be interested in seeing how well it actually reassembles all the backed-up chunks into the final file, and also if it’s possible to restore an entire hard disk from backup (presumably after booting from another hard disk or a utility device like Micromat’s TechTool Protege). After all, it’s not really the backup that’s important, but the restore.
[Jeff Carlson] Although I can’t yet see myself using CrashPlan because my Retrospect system works just fine for now, the product does seem like a good tool to install on Macs belonging to extended family members. I’ve helped my parents set up regular backups, but their systems so far haven’t included off-site backups; just getting files backed-up to an external hard drive has been a huge improvement. But CrashPlan could be a no-hassle solution for me to make sure, remotely, that their important data gets saved outside their homes, maybe even to a hard drive I set up in my office. Love means helping your less-technologically-inclined family maintain good backups.