GlusterFS vs Ceph, which is better for production use for the moment?

Question

I am evaluating GlusterFS and Ceph, seems Gluster is FUSE based which means it may be not as fast as Ceph. But looks like Gluster got a very friendly control panel and is ease to use.

Ceph was merged into linux kernel a few days ago and this indicates that it has much more potential energy and may be a good choice in the future.

I am wondering which(even out of the two?) is a better choice for production use? It would be nice if you could share your practical experiences

Solution 1

I know of a few GlusterFS installments being used in production. You may want to look at some of the Scalable Informatics products if you want a commercially supported system.

The GlusterFS team also had a booth at Supercomputing 09 where they demoed their systems. I think it's by no means a "mature" technology, but it's certainly usable in production in its current state. Note that by their very nature clustered filesystems are finicky to set up and maintain.

As far as Ceph goes, I've never heard of anyone using it for anything other than testing. It itself is based on a regular filesystem that hasn't even reached maturity yet (Btrfs). I wouldn't use it for anything other than a plaything at this point.

Solution 2

From the Ceph wiki at [ http://ceph.newdream.net/wiki/ ]:

"Ceph is under heavy development, and is not yet suitable for any uses other than benchmarking and review."

Also, Ceph does seem to make use of btrfs which is under active development as well. Between the two, GlusterFS should be the obvious choice for your needs as of now.

Comments

[Rahim]

Yeah, that pretty much makes it unsuitable for any production systems.

[Erik Aronesty]

That comment is outdated as of today. Ceph is now considered a production-ready filesystem by developers at Ceph.

[Lorin Hochstein]

@ErikAronesty CephFS still isn't considered production-ready by Ceph developers, although the object storage and block device interfaces of Ceph are considered production-ready. See sebastien-han.fr/blog/2012/12/03/ceph-and-mds