The best backup policy

To make sure you never lose essential data, consider implementing a 3-2-1 back-up approach:

  • You need to have 3 copies of your data. That means that next to the original file, there must be two backup copies.
  • The original files and one set of backup data can reside at the same location but they must be stored on 2 different media. This means you should not store a backup of a disk on that same drive. If the drive dies, you lose both the original files and the backup!
  • You must have 1 offsite set of backup data. This ensures that if a fire, theft or malware leads to the loss of the onsite data, there is still a spare copy at the other location. Many prefer a cloud based backup solution, like Carbonite, Crashplan or Backblaze. With these backup solutions, the data are stored in a datacenter, possibly in another state, country or continent. You can however also store hard drives with company data at home or keep a copy of data at your parent’s place.

There are a few other recommendations for ensuring your data are safe:

  • The two sets of backup data are ideally created using 2 different backup solutions or mechanisms. This ensures that if one application fails or stops working, the second backup mechanism still guarantees you have a full set of data.
  • From time to time you need to check your backups by restoring a few files from the backup. That way you know the backup is not corrupted and you remain familiar with the proper procedure to restore files.
  • Ideally one of the backups also contains versioning data, meaning that it is not just the current version of files that gets backed up, but also older versions. If a file accidentally gets corrupted (by bitrot, a software bug or another accident), your backups will also contain a corrupted file. That makes those backup worthless as well. If you still have an older version of that file or an older backup, that at least gives you a chance to recover the data.
  • At least one of the backup solutions must be fully automatic. Manually backing up data is perfectly possible but it is boring and can easily be delayed. Disaster will strike after backing up got postponed for months (or years) and the only backup you still have is completely outdated.
  • Different types of data may mandate a different backup strategy. Essential data must be backed up in real time or every night. Less critical data can be safeguarded every weekend. Many companies create a mirror of the startup disk of computers once or twice per year, just to avoid the hassle of having to reinstall the operating system and all the applications if a computer dies.

What to back up

There is no need to backup all the data on all the systems. There are thousands of files, such as temporary files, that are irrelevant. It pays to be selective in the files that get backed up.

  • Backups become easier to manage if all the important data used on all systems are centralized. Backing up one single data volume on a server or NAS system is easier than configuring a backup application on multiple systems. Before thinking about a backup strategy, first make sure there is an overall storage policy.
  • It is easy to overlook configuration data but often people spend a lot of time configuring InDesign preferences or creating Photoshop actions.

My own personal backup strategy

At home all essential data are centralized on an SSD on our main desktop PC. Windows 10 File History is used to back up those data, including older revisions of files,  to an internal hard disk. Synology Cloud Station Backup is used to create a second backup on a NAS box in the network. From time to time I manually copy the data to one of two external USB drives. During family visits, the drive with the most recent copy moves to a closet at my parent’s home.


5 January 2017

One response to “The best backup policy”

  1. James says:

    Triple redundancy has been recommended for decades for just about everything that is mission critical. It’s earliest computer related use was to parallel process three independent computers with the same data and have a fourth computer compare the results, flag the failed computer if one of the results does not match, and present the matched data as confirmed output.

Leave a Reply

Your email address will not be published. Required fields are marked *