Every piece of data on disk is mirrored using RAID-1. This means that for each disk drive, there is another drive of the same size copying everything the first drive is doing. Every single modification to disk is simultaneously performed on a backup disk. If one of those disks fails, the other one is still running to keep things going smoothly. On our server, there are two physical drives, behaving as a single logical drive.

Every three days, we do an incremental backup from the active partitions to a spare container. Monthly, we do a full backup of the system. This is helpful because accidentally deleted files can sometimes be recovered if they existed for more than a few days (though you shouldn't rely on this!).

Periodically, we take all of the online backups, encrypt them, and archive them on Amazon S3. This is a somewhat expensive process, since the backups are very large, so it doesn't happen as often as we'd like.

We also pay strict attention to security. Routine security audits are run frequently, unnecessary services are disabled, software patches are applied, and file permissions are checked often. The server itself is stored in a secure facility with limited access. We hope to continue our excellent security record, and we can, with your help. (See Section 3.4.1, “Choosing a password”.)