What is a backup?
In the simplest terms, a backup is simply a copy of one or more files. An example of a very simple backup would be making a copy of a file before making changes to it, so that you can go back to the original version if you don't like the changes. Backups can also become very complex, like a system that backs up hundreds of machines over a network to a backup server. The average person's backups should be somewhere in between.
The most important aspect of a backup is that it is reliable. An unreliable backup strategy often gives people a false sense of security. If you assume your backup is working, you're in for a bad surprise when you need it and find it isn't working. There are a number of different things you need to consider when creating your backup strategy:
Store on different media
I can't count the number of times I have seen people backing up files onto the same hard drive... basically, just copying them from one place to another on the drive. They feel safe because they have a backup. But what happens if the drive goes bad? Both copies are gone! Consider the following illustration:
Note that both files have been lost, since they were both stored on the same drive and that drive was damaged. Thus, you need to keep your backup on different media to ensure that there is always a good copy. For example, consider the illustration below in which the hard drive goes bad. Note that there is still a copy of the file!
Store in different places
Another common mistake is to back up to a hard drive sitting right next to the computer. While this is adequate protection against common causes of hard drive corruption, what happens if there is a fire? If the house is hit by lightning? If a burglar steals the machine and the drive connected to it? In all these cases, you've lost all your files!
So, what's the solution? Keep a copy of your backup in a safe place! If you have a safe, you could put backups inside the safe, but be aware that a fire-proof safe is only designed to keep paper documents, money, jewelry and other such things safe in a fire, and even then only for a limited time. The temperatures inside the safe may get high enough to damage CDs, DVDs or hard drives. You could buy a media safe, which is designed with digital media in mind, but they are expensive.
Another option would be to keep a copy off-site. You could take a backup to your office and keep it in your desk drawer. You could put a backup in a safe deposit box at the bank. You could leave one at a friend's house. Anything to put it safely in a different location, so that the same disaster can't affect both. (Well, unless a giant asteroid hits your town, but then you've probably got other concerns...)
You may also consider subscribing to some kind of internet-based backup service, such as Mozy. (Note that this is not necessarily an endorsement of Mozy, this is simply the most well-known service at the time of this writing.) Updating a backup on such a service is much slower than writing to an external hard drive, and you're trusting your data to someone else, but it does put the data safely out of your house without you needing to carry it anywhere.
Do it frequently
There is almost no such thing as backing up too frequently. You should back up at least once per day. Of course, this may not be strictly necessary... how often you need to back up depends on how much you're willing to lose. I, for example, often don't make what I consider irreplaceable changes on a daily basis. I usually back up about once per week unless I'm doing a lot of work or have done something I would really hate to lose (like load a bunch of photos into my Aperture library). I would not recommend this for everyone, and if you aren't backing up frequently enough, it is your responsibility if you lose data.
One backup is only somewhat better than none at all. It is not unheard-of for someone to need to erase their hard drive and restore from the backup only to find that something is wrong with the backup. If you end up in this situation, you will be very upset! Remember, a backup is no longer a backup if it's the only copy. You should keep a minimum of two backups so that, if something happens to one copy of your data, you still have two other copies.
Also, if you keep backups on physical media stored off-site, they will have to be in the presence of your computer for the backup. If you have to run to the bank, grab your hard drive, bring it home, do a backup, then run it back to the bank, you're probably not going to back up very often, which kind of ruins the whole point of backing up. With more than one backup, you could keep one backup hard drive on your desk at home and store one off-site at work or in the bank. As frequently as possible, swap them, so that the one stored off-site remains up-to-date.
There are many choices for the type of media you're storing backups on. Very old machines may still have floppy disk drives, but I can't recommend those, as the capacity is too low and the disks are too delicate. That leaves you with CDs, DVDs, hard drives and flash drives as reasonable choices.
CDs and DVDs are popular options, as they are cheap. Of course, they also represent a constant flow of cash, especially since the more expensive re-writable CDs and DVDs are less reliable and can't be re-written indefinitely. (Sooner or later, they wear out and will fail.)
Hard drives have far more storage capacity than CDs or DVDs, and they're getting very cheap these days. Hard drives have a cheaper price per gigabyte than decent-quality CDs or DVDs, meaning you get more for your money. They also don't represent a one-time-use investment... you can reuse them for quite a long time. Copying files to and from hard drives is also much faster with hard drives than with CDs or DVDs.
Flash drives (aka thumb drives or jump drives) can also be used for backups. Flash drives are fast and easy to use, and they tend to be more durable than hard drives. You can drop a flash drive on the floor and it'll probably be fine, while the same may not be true of a hard drive! The main drawbacks of flash drives are storage capacity, price per gigabyte and limited lifetime. Although price has been coming down in recent years, flash drives still cost far more per gigabyte than hard drives. Not only does this mean extra expense, it also means you won't find large capacities. An external 1 terabyte (TB) hard drive (equivalent to 1,024 GB) can be found for around $150, at the time of this writing, while the same price will get you 32 GB on a flash drive. However, if you're only backing up a small amount of data, buying a few cheap flash drives can be cheaper than buying hard drives. Of course, there's still that third drawback... a flash drive has a limited lifetime, especially if you're writing to it frequently. They will fail far sooner than a hard drive.
Tape drives are another possibility, but they are expensive and have other drawbacks. They are more often used with large backup systems, and do have the advantage of making off-site storage and archival of backups relatively easy. However, since the expense puts them out of the running for the average person, and since I don't have much recent experience with tapes, I will not discuss them in further detail.
What about archiving data?
Archiving is a specific kind of backup. With most backups, you'll recycle media over and over, keeping only the most recent copies of your data. The idea behind archiving is that you keep the data permanently. "Permanently" implies a very lengthy time period over which to keep data intact. If you are going to keep data for a very long time, you need to think carefully about media type and quality and you need to think about methods of storage.
For archives, you can't use any rewritable CDs or DVDs (CD-RW, DVD-RW and DVD+RW). The media is designed to be changed, so you can't rely on it sitting on a shelf for decades without changing! You also need to be careful about the brand, as some brands are better than others. Some, you may be lucky to get ten years out of. For more information, there is a good article on Ad Terras Per Aspera that describes the technical issues and gives recommendations for archival-quality media. Of course, regardless of media quality, if you're going for true archival, I'd recommend having at least two copies of each archival disk, and would also recommend checking the integrity of the disks every few years.
Hard drives can easily be used for archival purposes, since they tend to be very stable storage. It may seem like a waste to put a hard drive on a shelf an not touch it again, but keep in mind that, if you are storing large amounts of data, this actually costs you less per gigabyte than doing the same thing with CDs or DVDs.
You can further reduce cost by buying hard drives without an enclosure. An enclosure is a shell you put a hard drive into in order to use it as an external drive. External hard drives come inside an enclosure, while drives bought as "internal" drives, meant to go inside a computer, do not come with this enclosure. This enclosure costs money, so a drive costs more as an external drive than as an internal drive. You would put them into an external case to use them, and then take them out again to archive them, meaning you don't have to pay for enclosures to sit on the shelf. Some manufacturers even make special swappable drives, similar to the old removable Zip drives that were so popular in the 90s.
Of course, as with archives on CDs or DVDs, you should keep multiple copies of each set of archived data and should check the integrity of your archived hard drives every few years. If a drive fails, you need to be prepared to make a fresh copy on a new drive from a second copy.
Flash drives, unfortunately, don't make good archival media. As mentioned previously, they fail faster than hard drives. Flash drives are ideal for cheap and easy storage of small amounts of data, but they do not make good archives. If you need to archive small amounts of data, use high-quality CDs or DVDs instead.
There is also more to archiving than choice of media. The best place to store archival data is in a location that is cool, dry, dark and dust-free. For magnetic media, such as hard drives, it is also a good idea to avoid strong electrical or magnetic fields.
What kind of backup do I need?
There are several different kinds of backup. The problem with describing them is that I've seen different backup programs use the same terminology in different ways. So if you decide that you like what I call "incremental" and use that option in your backup software, it may not do what you think. Be sure to look at the documentation for your backup software to be sure you're getting the backup that you want.
Obviously, a backup must store everything that you tell it to. If it doesn't, it's not much use. There are, however, different ways to achieve this goal. For most backup software, the first time you perform a backup is the same: everything from the source (the set of files you have said need to be backed up) gets copied into the backup. This is referred to as a full backup.
After the first backup, types of backups can vary widely. One choice you could make would be to perform a full backup each and every time. This is not usually advisable, though. If you are doing your backups frequently enough, most of the files in the source won't actually have changed since the last backup, so copying them again is a waste of resources.
The next choice is something like a differential or incremental backup. This is where terminology starts to get fuzzy, as I have seen programs that use one of these terms in the same way that another program uses the other. The basic idea of both, however, is that only the data that has changed is copied. The advantage is that less data needs to be copied. The main disadvantage - that your backups end up spread over and relying on multiple separate media - primarily applies to backups stored on tapes, CDs, DVDs or other such media. If you are backing up to a large-enough hard drive, this does not need to be an issue.
In addition to the issue of what files from the source get backed up, there is also the question of how many versions are kept. In some backups, if a file is already in the backup, and that file changes and is backed up again, the new version of the file replaces the older version in the backup. Thus, there is only ever one version (the newest) of any file in the backup. This kind of backup is often referred to as a clone, since the contents of the backup should be an exact copy of the source. If the clone is created in such a way that you can start up the computer from the backup, it is called a bootable clone. (More on this shortly...)
Other backups will keep both versions of the file, allowing you to restore not only the newest version of the file, but any older versions as well. This can be an extremely powerful feature of your backups, as it allows you to reverse changes you make to a file. It also provides extra insurance against file corruption. You might not notice that a file has been damaged for some time, and if only the newest version of the file is kept in your backup, you will have lost the data, since the backup will contain the corrupted file. The main drawback of this kind of backup is that it requires more storage space (potentially quite a lot more space) than is required by all the source files together.
Finally, some backup software will back up not an entire file that has changed, but only the specific bytes within the file that have changed. This saves a lot in storage space, but in my opinion does so at the expense of reliability. I tend to prefer backups that copy the source files themselves into the backup, so that I can always access the files without any backup software at all. If your backup software copies files into some kind of archive file or files, you are forever tied to that software. (Okay, maybe not forever... just for as long as you want to be able to access the data stored in it.) If the company goes out of business and stops updating the program, and then it breaks without warning after some future system update, you're up a creek without a paddle! There will be no way to access your backups without downgrading to a previous version of the system. In addition, if your backup is stored in one large file, and one small part of that file becomes damaged, you may lose the entire backup. Be very cautious with byte-level incremental backups!
Do I need a bootable backup?
The bootable clone backups mentioned a few paragraphs above have some very distinct advantages and disadvantages. One major advantage is that a bootable clone allows you to get back up and running very quickly after a disaster. Unfortunately, this can also be a disadvantage... an un-savvy user may start using the system and data on their only backup, thus putting the only copy of the data at risk! Thus, it's important to either have multiple backups (as mentioned previously) or to copy the clone back to your computer as quickly as possible without working from the backup.
Another major disadvantage of bootable clones is that they include your entire system, all your applications, etc. This uses up a lot of space in your backup storing things that can easily (though less quickly) be restored from the original disks. In addition, if you lost data due to a damaged system, your clone very well may have the same damage in its backed-up copy of the system, so often reinstalling is preferable. On the other hand, if your hard drive died and your system was just fine before, it's much quicker to just copy the clone back to a new hard drive than to spend time reinstalling everything.
Okay, I've chosen one type of backup. What next?
Choose another! No, I don't mean that you've chosen wrong... I'm not a mind-reader, after all. What I mean is, in the spirit of using multiple backups, you should consider using two or more different backup programs. Remember that backup software is still just software, and as such, can have bugs. If your only backups are made with one particular program and it has a bug that causes you to lose data from your backups, you'll be glad if you had a second set of backups made with another program! Similarly, it would be a good idea to choose a couple different types of backup... for example, you might keep one bootable clone, made with one program, and a couple smaller incremental backups of only your user folder, made with another program.