Was in the same lab as Sage in grad school, and shared an advisor (Scott Brandt). He is super hard working, and definitely deserves this. Ceph was over 10 years of his life.
Fun fact - Sage was also the founder of webring, and simultaneously one of the dream host founders.
I would love to hear people's opinion's on what this means for ceph going forward? Will it change anything?
It's telling for me that as the project starts becoming 'hot' RedHat step in. I totally get why it's a great strategic purchase in the context of winning cloud adoption / share / love.
IMHO, the biggest thing will be that Ceph is now a product that is supported by a huge publicly traded company. That's a big deal and I'll explain why:
Storage is precious data is a big responsibility within most IT orgs. There are lots of regulations and best practices when it comes to protecting, backup up, verifying, and making available business records that are stored digitally. A Fortune 500 company isn't going to trust their business to a home grown solution (think self-supported Ceph compiled from source) when they can pay experts to do that work.
So, in the marketplace, Ceph competes against Scale-IO (now EMC) and Cleversafe. Scale-IO is on CTO radar because it's supported and sold by EMC. Even if Ceph is FREE it doesn't matter because you need support for these products to be actually deployable.
SO...I'm sure that the Inktank acquisition was predicated on a couple of things:
1. Potential customers saying that they liked Ceph but had problems getting it past management because Inktank was too small.
2. RedHat understanding (and verifying) that there is a huge opportunity in storage management and they were tired of getting frozen out by other solutions that were better supported.
I'm more interested in what it means for Gluster. I've had RH pushing Gluster at me very hard as performance-competitive with Ceph. Now they have Gluster and Ceph, it'll be interesting to see if Gluster falls by the wayside, or if they reposition Gluster more as easily-configured resilient storage.
CephFS (which is a distributed file system) is orthogonal to XFS or BTRFS which are local file systems. Since Gluster is providing a file system solution at this point, that may be the reason for the de-emphasis. However, Gluster is not a parallel file system, and Ceph is much better suited to fill that role. So I suspect CephFS will continue to be developed, if anything to compete against Lustre, PanFS, GPFS, etc..
Why do you say that GlusterFS is not a parallel file system? Certainly it is by the common definition of "file system that spreads data across multiple storage nodes."
That's because (a) CephFS isn't ready for prime-time (according to every talk I've seen at linux.conf.au in the last two years) and (b) the early push for btrfs as the underlying storage for Ceph has faded in favour of XFS until they're happier with the state of btrfs.
I would really appreciate if somebody can clear this up for me:
Can one mount Ceph on multiple nodes? (I'm referring to the CephFS part).
I've worked with OCFS2 (which allows you to do exactly this) and liked it very much, but would've loved a clustered backing storage. Which is what Ceph provides.
I think this is a clear play at driving some of the key components in the OpenStack world, with GlusterFS (subject of an earlier Red Hat acquisition) and Ceph being the two main distributed storage choices.
Fun fact - Sage was also the founder of webring, and simultaneously one of the dream host founders.