The main purpose of a distributed filesystem is to allow the creation
of folders that contain more data than a RAID of a single computer is able to
store. Redundance is not the primary requirement, only a mean for
increasing the reliability of a distributed filesystem instance. It is OK,
even preferential, for the distributed filesystem implementation to use
unencrypted connections between computers, because if the encryption of the
connections is left to
encrypted
tunnel implementations like the SSH
tunnels, then it is possible to upgrade the encryption software without having
to modify the distributed filesystem implementation.
- BeeGFS (Parallel Cluster File System)
- CephFS is POSIX compliant
- Coda
- Gluster
- LizardFS
- Lustre was developed for scientific compute clusters by the Sun Microsystems.
- MooseFS
- OpenAFS
- tahoe-lafs.org
- XtreemFS related discussion thread(archival copy)