Cluster file system GlusterFS 2.0 released
The developers of the cluster file system GlusterFS, not to be confused with the now Sun-owned ClusterFS company that make the Lustre cluster file system, have released version 2.0 of their software. GlusterFS is able to aggreagate various storage systems connected over Infiband or TCP/IP to create one large parallel network file system. GlusterFS uses FUSE and runs in the userspace, making it more portable and easier to install. It has been deployed in production with a number of organisations.
New in version 2.0 are improvements to the atomicity of write operations to give better reliability with hardware failures and an automatic high-availability/fail-over mechanism, which will seamlessly switch over in the case of an interface failing. A distributed hash table has now been used to improve GlusterFS performance when dealing with small files. Applications now also have the option of bypassing the FUSE layer and using libglusterfsclient to talk directly to the GlusterFS process. The software is available to download for Linux (CentOS, Debian, Fedora, OpenSUSE, RHEL, SLES and Ubuntu), Mac OS X and FreeBSD, with a Solaris port in development. It is licensed under the GPL3, and Gluster.com offer a subscription based support package for the production user.