Keep copies files and directories for backup purposes by creating a new directory tree each time it is used, storing all specified data there. Files that did not change since the previous backup are linked, not copied, thus speeding up the process and saving much space, because in typical environments, only few files change between different runs.
As a result, often you can have daily full backups for weeks and months online for instant access. Restoring files from them requires no software. It suffices to backup and restore the whole tree for disaster recovery, file based backup is not required. Alternatively, use NFS as destination and revoke write access afterwards.
Keep can employ multiple processes to make use of I/O subsystems that allow multiple simultaneous operations.
The source is available as a GNU zipped tape archive.