Chapter 7. Open-Source Near-CDP

Chapter 8 discusses commercial backup software and the concept of near-continu ous data protection (near-CDP) systems, which is basically replication tied with some type of snapshot product. Before moving to the commercial side of things, this chapter examines three open-source near-CDP systems. The first section covers rsync with snapshots and explains the concept, originally popularized by Mike Rubel. The rsnapshot section explains an open-source project based on that concept. It's designed to automate the backup of user files and does not work very well with databases and other large files. Finally, the rdiff-backup section explains a different project with a number of features, including advanced metadata capabilities and an ability to handle large files.

This chapter was contributed by Michael Rubel, David Cantrell, and Ben Escoto. Mike is a graduate student in aeronautics at Caltech, where he keeps several backups of his thesis (which he hopes to finish soon). David doesn't believe in quiche. Ben is currently an actuarial analyst at Aon Re.

Replication by itself wouldn't work. You can't just have one system rsync to another system as a backup because logical corruption (such as a virus or user error) would be replicated to the backup. You need to have some way of keeping multiple versions of data on the target so that you can go back to a previous version when necessary.

If you can do this, you can have a continuously incremental system that always has a full backup of any version available. This requires, of course, enough space to store one full copy plus incremental backups. It can even be employed in a variety of ways: push-to-server (as just described), pull-from-source, or locally (for example, to a removable hard disk drive).