Some backup programs distinguish between incremental and differential archiving schemes. Although not all software uses the terms in exactly the same way, the difference is typically that in an incremental backup, only the files changed or added since the last time the backup ran are added to the archive. With a differential backup, all the files changed or added since the initial full backup are added to the archive. Thus, differential backups take longer to run than incremental backups.
This distinction is important when backing up to tapes or other removable media, because it affects the speed with which a backup can be restored. When restoring from an incremental backup, the software must copy the entire initial backup and then step through each of the incremental backups to retrieve all the updated files. This can require a great deal of media swapping. A differential backup, on the other hand, can be restored more quickly because the software must copy only the original backup and the most recent one.
When backing up to a hard drive, however, this distinction is less significant, because the random-access nature of a hard drive enables it to restore either sort of backup with roughly equal speed.
One consideration in choosing a backup schedule is media management. For example, if you're backing up to a recordable DVD, you must be prepared to insert a blank disc whenever the schedule runs. Swapping media can be an intrusion into your normal routine (especially if that routine involves the frequent use of other discs in the drive you use for backups). On the other hand, if you schedule backups to run when you're not around, you must always think ahead and make sure the drive has the necessary media ready. If, on the other hand, you're backing up to a hard disk or network device that can stay connected all the time, this problem occurs less frequently, if at all.
Depending on the speed of your computer, which software you use, and how you configure it, you may find that your computer slows down significantly while backups are running. This could be an argument for scheduling backups for when you're not using the machine. However, if you do not leave your computer on all the time, you will need to take special care to ensure that it's on and ready when the backups are scheduled to run.
How often should you back up your computer? And if you're making both duplicates and archives, how often should you update each?
No single answer is right for everyone, but as a starting point, my rule of thumb is that duplicates should be updated at least as frequently as major changes to your system (such as installing Mac OS X updates or new versions of applications), and archives should be updated every day you make minor changes (receiving email, modifying text documents, and so on). Thus, if you use your computer heavily every day, and often install new or updated software, you might opt for weekly updates of your duplicates and daily updates of your archives. On the other hand, if you use your computer only occasionally, the schedule could become once a month for duplicates and once or twice a week for archives. Under no circumstances do I suggest backing up less frequently than once a month or more frequently than twice a daythe risk is too high in the former case and the aggravation too great in the latter.
Tip:Always update your duplicate just before installing system software updates. That way, if the new version of the software contains any serious problems, you can easily roll back your system to its previous state.
There may be some cases in which you could not afford to lose even a half day's work in the event of a serious problem, making twice-daily archives seem inadequate. If you're working on an important document, there's nothing wrong with copying it to another volume once per hour or as often as you feel it's necessaryor scheduling your backup software or a synchronization utility to do so for you. But updating an entire archive that frequently is likely to slow your work.
Keeping Multiple Backups
A sound backup strategy always includes backups of your backups! Picture this: you've diligently backed up your computer's internal hard disk to an external drive. Then one day, lightning strikes and both drives are damagedor your home is robbed and all your equipment stolen. So much for your backup. Backup media can fail for all the same reasons your hard drive can fail. So having just one backup, in my opinion, is never enough. You should alternate between two or more sets of backup media for greater safety. If you've set up your backups to run on a schedule, this might mean using set A (a hard drive or a stack of CDs) every day for a week, then switching to set B (a different drive or stack of CDs) for each day of the following week, then switching backand so on.
So are two sets enough? It depends. Most experts recommend using at least three sets, of which one is always stored off-site. But this advice was first given in the days when the media commonly used for backups was much less reliable than what's available today. And the cost of three sets of mediaespecially hard drivescan be hard to swallow for the average home or small-business Mac user.
In my opinion, except for mission-critical business use, two sets each of duplicates and archives should be adequate for most users. If you back up to hard drives, this can mean two drives, each of which is partitioned to store both a duplicate and an archive. Of course, if you can afford a third set, your data will be somewhat saferand your backup routine will be somewhat easier. In any case, you certainly should keep one of those sets in another location all the time.
Backing Up a Small Network.
To this point, I've assumed that you're backing up a single Mac. But what if you have several in your home or office? How does this affect your backup strategy?
One approach is to back up each machine separately. This may involve keeping separate stacks of recordable CDs or DVDs next to each machine, or hooking up external FireWire drives to each one (though you could, of course, move a single high-capacity drive from one computer to the next). If your backup needs are relatively small, there's nothing wrong with this approach. But if you have more than a couple of machinesespecially if their hard disks contain a lot of data that you can't afford to losea wiser strategy would be to back them all up at the same time over your network.
You do have a network, right? If you have multiple machines that aren't currently connected (whether by Ethernet cabling or AirPort wireless networking), you should hook them up. Not only does a network enable better backups, it makes transferring files and accessing the Internet much easier.
Network Backup Approaches:In a network backup, one computer functions as the backup server. This is the machine to which your backup device(s) are physically connected. Files from your other machines are copied over the network onto each backup device. Network backups can proceed by three different methods:
The server shares its backup volume (using AFP, FTP, or SMB; see the Glossary for info on these acronyms), which the client machines mount as a volume in the Finder. Then each client machine uses its own backup application to back up files to the network volume (rather than a locally attached hard drive or optical drive). This is sometimes called a push backup, as each client "pushes" its data onto the network volume.
- Each client machine shares its hard disk (again, using AFP, FTP, or SMB). The server mounts each of these volumes in the Finder, and then the single copy of the backup application running on the server copies files from each of the network volumes onto its locally attached backup volume. This is sometimes called a pull backup, as the server "pulls" data from each of the clients onto its backup volumes.
- The server runs backup software that supports client-server network backups, and the other machines run client software that communicates with the server directlywithout any of the machines having to share or mount volumes.
Some SMB servers limit the size of any single file to 2 GB; others limit it to 4 GB; still others have limits as high as 2 TB. Because some backup software transmits all your data over the network as a single file, you may run into situations where you cannot back up more than 2 GB (or 4 GB) of data to an SMB server. If you can't persuade your system administrator to update the server software to a version that supports larger file sizes, you may need to use a different server (or different backup software).
Almost all backup applications support push and pull network backups, but I recommend against them. For one thing, network volumes can become disconnected for any number of reasons, and if a volume is unavailable when it's time for a scheduled backup, that backup will fail. A few applications can try to mount missing volumes for you (even remembering user names and passwords, if necessary), but even this is no guarantee of success. Push and pull backups are also inherently less secure than client-server backups, and are sometimes quite slow. Also, in the case of pull backups, file ownership may change in unacceptable ways, making bootable backups impossible. Sometimes push backups can be bootable, but it's a dicey operation.
True client-server backups require less effort, are more secure, and tend to offer more flexibility. Often, client-server backup software also supports multiple platforms. Of the backup software covered in this article, Retrospect, RsyncX, and BackupSW offer client-server backups. Retrospect and BackupSW both support Mac OS X and Windows; Retrospect also supports Mac OS 9, while BackupSW also supports Linux.
If you need to back up a small Macintosh or Macintosh/Windows network, I recommend Retrospect Desktop, which includes a license to back up the machine on which it's installed, plus two more client computers (additional client licenses are available at $37 each, with volume discounts if purchased in packs of 5, 10, 50, or 100). You'll get the best results with the Backup Server script, using hard disks that are large enough for all the data on all the Macs.
Besides selecting the right software, several other matters require your attention when planning a network backup system:
- Media: Although optical media or other removable storage may be acceptable for single-machine backups, for best results, network backups require storage devices that are always available. In other words, hard drives are the best bet for small networks. Also, if you're making duplicates that you may later wish to boot from, be sure to partition the hard disks in such a way that each startup disk on the network gets its own partition for a duplicate.
- Bandwidth: You can perform a network backup using an AirPort wireless network, but even with AirPort Extreme, you get only a small percentage of the bandwidth that a wired 100Base-T Ethernet connection will give youso backups will take much longer, especially if you're duplicating an entire hard disk. In any case, you definitely want the highest-bandwidth network connection you can get. If your computer uses multiple network interfaces, open System Preferences, go to the Network pane, and choose Network Port Configurations from the Show pop-up menu. In the list that appears, drag Built-in Ethernet to the top and click Apply Now to ensure that the wired network is used in preference to AirPort when both are available.
Note:Every network is different, but I have seen cases where Retrospect client-server backups are unreliable when the client machines' IP addresses are dynamically assigned by an AirPort base station. If this happens to you, consider assigning (private) static IP addresses to each client.
- Availability: For a scheduled network backup to occur, both server and client machines must be turned on and awake. If your machines are currently not left on all the time, check the Energy Saver pane in System Preferences on each computer to ensure that it will not be off or asleep when backups occur.
Tip:Scheduling network backups for times when all machines are available can be a challengeparticularly if you have laptop computers that are not always on the network. Retrospect offers a great feature called Backup Server that constantly polls all the clients on a network. If it sees one that hasn't been backed up in at least 24 hours
Remote Backups:In the discussion so far, I've assumed that the machines you need to back up are connected to the same local network as your backup server. But what if you travel frequently with a laptop? Can you use a broadband connection at a hotel or Internet café to copy the files to your server over the Internet? The short answer is: Maybe.
"Push" backups work only if you can mount your backup server's volumes remotely; "pull" backups work only if your server can mount your laptop's volume remotely. Sometimes this works, but often notyour firewall at home must enable access to the necessary ports, and the ISP providing your remote access must also permit file-sharing access over their network. You also run a certain risk that your files may be intercepted in transit by a hacker, unless you take extra steps to encrypt the network link between your laptop and your server.
Client-server backup software, such as Retrospect, normally polls only the local network for available clients. In some casesfor example, with the more-expensive Retrospect Workgroup or Retrospect Server packagesyou can manually enter an IP address for a computer outside your local network. However, if you're traveling and don't know what IP address you'll have at any given time, this method is problematic. One possible solution is to use a dynamic DNS service, such as the one provided by easyDNS (www.easydns.com/dynamicdns.php3), to assign your laptop a domain name whose IP address changes as needed, and then enter that domain name in Retrospect.
This problem is more readily solvable using a VPN (virtual private network) connection to your home network, but the details of setting up such a system go beyond what I can cover in this article. As a lower-tech workaround, consider packing some DVD-R media for temporary backups when you're on the roadand be sure to store the discs separately from your laptop!
What I recommend for most users is a two-pronged approach: periodically scheduled (say, weekly) duplicates of your entire hard disk, and even more frequent (say, daily) archives of your data files.
The duplicates will provide you with a complete, bootable copy of your hard disk, while the archives will pick up all the files that change regularly. Users with extensive photo or video data may need to go a step or two furtherseparating that data from their main backups and using special strategies to keep it safe without incurring enormous media and equipment expenses.
You should create duplicates (onto hard drives, ideally) of your primary disk and any other startup volume you normally use. If you have a single, unpartitioned hard disk, then you have only a single volume to worry about. If you have multiple partitions (or multiple internal or external hard drives) that contain bootable systems, I recommend making duplicates of all of them. If a hard drive fails, after all, it can take with it all the partitions it contains; and a disaster that wipes out a single drive could wipe out all of your drives.
Note:When you create a duplicate, you copy everything from the source drive to the target driveincluding, of course, all the files that make up Mac OS X. Therefore, there is no need to install Mac OS X on your external drive before creating a duplicate.
Most duplication software enables you to deselect individual folders you wish to exclude from a duplicate; some use selectors, exclusions, or both. Although you could make an argument that some files are not worth including in a duplicate (such as the cache files located in ~/Library/Caches), the safest and most reliable tactic is simply to include everything. A file or folder that seems irrelevant to you may turn out to be crucial to the functioning of your system.
Archive Strategy:The archives you create should include all your important files (on each volume you use regularly, if you use more than one). The main question, though, is how you determine which files those are.
Some people suggest performing a full archivethat is, archiving every single file on your disk, just as you do when creating a duplicate. Others suggest performing a selective archive that includes only user-created data files, especially those that change frequently.
With a full archive, you have yet another copy of all your files besides your duplicatesan extra insurance policy. Restoring a full archive to an empty disk requires fewer steps, and less time, than restoring a selective archive (since in the latter case, you must restore a duplicate first). On the other hand, a full archive requires significantly more storage space, increasing your media cost, and takes longer to run. In addition, some backup software does not enable you to restore an archive as a bootable volume. My own preference is for selective archives, though I would not discourage you from performing a full archive if resources permit.
If you do choose to archive selectively, a good starting place is your home folder. By default, this folder contains most of your preference files, the files shown on your Desktop, and data for many of Apple's applications (Address Book, iCal, iTunes, iPhoto, Mail, Safari, and so on), among others. Although you can organize your hard disk however you want, Apple encourages you to keep all your user-created documents in the ~/Documents folder or elsewhere in your home folder. So it could be that all your important, user-specific data files exist somewhere inside your home folderand if not, presumably you are aware of the locations of folders you've created elsewhere.
But even if you have assiduously colored within the lines and kept all your personal data in your home folder, should you archive the whole thing? In some cases, the answer is no.
Because Apple designed the home folder as a catch-all, it has the tendency to swell to enormous sizes. For example, if you maintain the default settings in iDVD, iMovie HD, iPhoto, and iTunes, all your digital media will be stored in your home folder. If, like me, you've imported your entire collection of CDs into iTunes, you may be looking at a huge Music folder (mine is well over 20 GB, and that is small compared to some). If you store digital video on your computer, your Movies folder will certainly be even larger.
Although there's nothing wrong with adding all those files to your archive, it may not be strictly necessary eitherbecause all those files should already be backed up safely as part of the duplicates you maintain. If, as in the case of imported CD tracks, digital photos, or video downloads, you modify those folders less frequently than you perform duplicates, you might consider saving time and space by excluding them from archives. But if in doubtespecially when it comes to irreplaceable photos and videoerr on the side of including them; having an extra backup just may save your bacon one day. Purchases from the iTunes Music Store also require special handling as I describe next.
Besides digital media, you may wish to manually exclude certain other files from an archive, if needed to save space. For instance:
- Downloads: Applications and other files you've downloaded from the Internet can nearly always be downloaded again. It may not be worth dedicating significant media space to hold such files.
- Cache files: Temporary cache files, such as those stored in ~/Library/Caches, are not crucial to an archive, as they will be recreated automatically if needed.