Comparisons to other backup systems

Within this document, encrypted and on-line are used in very specific ways to describe backup systems.

encrypted: The data to be backed up is transmitted and stored on the server so that the data can only be recovered if secret encryption keys are available.
on-line: The backup server does not use tape or any other removable archive media. All data is stored within the server file system on conventional hard discs, and every (current) file is always available.

Only backup systems which are both encrypted and on-line are compared in depth. Other systems have brief notes at the end of this document.

Please email me with any systems I have missed, or corrections to my analysis. This is not intended to promote Box Backup, but as an objective comparison to allow you to decide which system to adopt.

Encrypted, on-line systems

	Box Backup	duplicity	hdup
Development status	stable, under development, not feature complete	"not stable yet" (details)	stable, maintainance only
Changes only when uploading new versions of existing files (like rsync)	yes	yes	no
Connects to server using	TLS (SSL)	ssh/scp, local file access, rsync, ftp	ssh, local files
Authentication	SSL certificates (client and server)	Traditional file storage accounts on server (eg unix username, virtual FTP etc)	Traditional file storage accounts on server (eg unix username, virtual FTP etc)
Dependencies	OpenSSL	Python installation, librsync, GnuPG, remote shell	ssh, mcrypt/GPG, GNU tar
Signed data storage	Implicit	Explicit with GnuPG	Implicit/Explicit with GnuPG
Scheduling	Self-contained with randomised element to avoid cyclic server loading, or cron job	cron job	cron job
Continuous backup	Yes -- backs up files a configured time after being changed, or... No -- in snapshot mode	No -- takes a snapshot of the filesystem when it is run	No -- takes a snapshot of the filesystem when it is run
Handling of old versions and deleted files	Old and deleted files remain on server until store reaches a set limit, after which they are removed. No particular limits on availiability apart from this size limit.	Full and incremental backups, deletion by user scripts	Monthly full backups, weekly and daily incremental, deletion of archives by user scripts
Resources used on client	Few Mb of memory, very small amount of disc space	Status and checksums for all files backed up, plus potentially large temporary files. (however, disc space is cheap, so this is not an issue unless the clients are older machines with small hard drives)	Minimal (latest versions can stream archives directly to server)
Resources used on server	Disc space	Disc space, remote shell server, UNIX accounts	Disc space, ssh, UNIX accounts
Redundant storage on server	Optional built in userland RAID storage	Relies on OS to provide RAID	Relies on OS to provide RAID
Restore	FTP-like client, or simple command to restore entire directories	Simple command to restore entire directories, fetch files or directories by name.	Simple command to restore entire directories, fetch files or directories by name, will fetch entire archives for single files.

Other systems

	Encrypted?	On-line?	Network	Notes
Amanda	no	no	LAN	The classic network backup system
Bacula	no	yes	LAN	Comprehensive job control. Stores lists of backed up files in a database for faster retrieval. Can backup to disk as well as tape. Supports encrypted transfer, but not storage.
rdiff-backup	no	yes	Remote	"The idea is to combine the best features of a mirror and an incremental backup."
rsnapshot	no	yes	Remote	Perl script. Pulls files over ssh from the source server to the backup server. Creates efficient snapshots of the source filesystem.
rsync	no	yes	Remote	Requires extra scripts for automatic scheduling and old version archival. May be less efficient than Box Backup, as it needs to check for existence of every file on the remote side, whereas Box Backup keeps an efficient record.
Mondo Rescue	no	no	n/a	Backup to local media, aimed at full system recovery.

"Network" is marked as LAN if the system uses so much bandwidth it can only be used on a LAN, Remote if it uses bandwidth conservation techniques which make it suitable for use over an internet connection.