What's an incremental backup?

A full backup is just that: full. It's a backup of everything. A "full backup of C:" for example would be a backup of all the files folders and contents of the C: drive. A full backup of "your system" might include the entire contents of all the drives connected to your system.

The important thing is that a full backup stands alone; it's a snapshot of your machine at a particular point in time. You might choose to restore files from that snapshot, or even restore the entire machine to the exact state it was in at the time that snapshot was taken.

Pro: a full backup stands alone; it has everything.

Con: a full backup can be big; it has everything.

An incremental backup includes only those things that have changed since the previous backup and saves those things into a separate, additional, backup file or location. By definition, the first incremental backup is a full backup; it backs up everything since there's no "previous backup" to compare to. The second incremental backup backs up only those files that have changed since the previous backup was taken.

Since the vast majority of files on your machine actually never or rarely change, an incremental backup can result in a much smaller backup. On my machine, a full backup is about 50 gigabytes, but the incrementals vary from as little as 2.5 gigabytes to around 10, depending on how much happened on the machine that day.

The cost of using incremental backups is one of management. Since each incremental backup relies on the backup that preceded it, in order to restore a machine to an arbitrary point in time all the incremental backups must be available to perform the restore. For example, I perform a monthly full backup and then a daily incremental. If I want to restore my machine to a point in time 10 days into that month, then I need the initial full backup plus the 10 incremental backups to reach that day. Each incremental backup must be restored "on top of" the backup prior until the 10th has been processed.

The good news is that most backup software actually makes this very easy, almost transparent. Typically, you point the software at the collection of backups - both full and incremental - and tell it "I want to restore to this date" and the complexity is managed for you.

Pro: significantly less disk space used compared to an equivalent set of full backups.

Con: the baseline full backup and all the incremental backups must be preserved and available in order to restore.

Differential backups are a kind of hybrid. In reality they're just incremental backups, but with a fixed starting point. Rather than backing up only the changes from the previous day, each differential backup includes all the changes from the baseline full backup. So day 2's backup would include changes from day 1 to day 2. Day 3's backup would include changes from day 1 to day 3.

Pro: Each day can be restored from only two backups, the initial full plus that day's differential.

Con: Differentials only to grow in size from day to day.

•

Let's compare backing up a "C:" drive. This table shows what each type of backup would include on a series of successive days:

Now let's look at restoring data. This tables shows which of the backups above must be available in order to restore to a specific day:

•

The choice of what kind of backup to use can seem somewhat daunting. Daily full backup seem like the simplest, and yet you'll quickly find yourself running out of disk space. Incrementals seem like a great alternative, but you could find yourself in backup management hell as you try to keep track of which files you need to keep for how long. And differentials seem like an interesting idea, but then are they really all that different than incrementals?

Here's my suggestion:

• Periodic full backups. I do a full backup once a month. I actually save these full backups for "a while", so that if I need to find something from, say, three months ago, I might be able to recover it from an old backup.

• More frequent incremental backups. I do a nightly incremental backup. As the month goes on, I accumulate a collection of incremental backups in addition to that monthly full that will allow me to revert or recover from any day in the month. Once a month I "clean the slate" as a new full backup is taken.

Why not just do a nightly full backup and allow it to overwrite the previous day's backup? You could, but ... you're assuming that you'll always detect a problem or a missing file within a day, before the backup containing it is overwritten. Experience shows that things don't work that cleanly. It's not at all uncommon to wonder days, sometimes months later, "where did that file go"? Being able to search your daily backups for a file that might have disappeared can often be a real lifesaver.

•

Finally, don't let all this intimidate you into not backing up at all. Most backup programs make these choices fairly easy, and will do a good job of managing the backup files for you. In my case, for example, after configuring Acronis TrueImage Home, I actually don't even think about my backups more than about once a month, and even then it's only to decide if I want to save the most recent monthly full backup for posterity.

Other than that, everything's completely automated.

Computers are good at that.