[OpenAFS] OpenAFS vs RedHat GFS

Russ Allbery rra@stanford.edu
Thu, 24 Aug 2006 09:41:04 -0700

Michael D Norwick <mnorwick@centurytel.net> writes:

> Could someone please explain the fundamental differences between OpenAFS
> and RedHat's GFS.  In reading their (RedHat's) list of features it
> sounds a lot like OpenAFS with changes in nomenclature.  They talk about
> charateristics such as scalable volumes, and fail-safe replication and
> I'm wondering if it's a difference in implementation of the same
> principle.

It's really a completely different thing entirely.  It's a difference of

GFS is fundamentally a clustered file system designed to manage
simultaneous access to the same disk among closely connected nodes.  It
requires that all the systems accessing the file system have access to the
same physical disk.  It is *not* a traditional distributed file system; it
has much better performance than one, but will not scale outside of a
closely-connected set of systems all sharing the same disk.

AFS is a global distributed file system that really comes into its own at
large sites with thousands of clients and multiple file servers.  It
supports user-transparent volume moves, space partitioning, network
authentication, transparent addition and removal of file servers, unified
namespace, and scalability to tens or hundreds of thousands of
heterogeneous clients on different operating systems.  It's mostly used as
a site-wide distributed file system.

Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>