Red Hat Acquires InkTank, Ceph Maintainers

An anonymous reader writes "Red Hat announced their pending acquisition of Inktank this morning. Sage Weil and a team of researchers at University of California Santa Cruz first published the architecture in 2007. Sage joined DreamHost after college and continued development on Ceph until DreamHost spun off a Inktank, a company focused solely on Ceph. In Sage's blog post on the acquisition, he says 'In particular, joining forces with the Red Hat team will improve our ability to address problems at all layers of the storage stack, including in the kernel.' Sage goes on to announce that Inktank's proprietary management tools for Ceph will now be open sourced, citing Red Hat's pure open source development and business models.

Ceph has seen wide adoption in OpenStack customer deployments, alongside Red Hat's existing Gluster system." Ceph looks pretty cool if you're doing serious storage: CERN has a 3 Petabyte "prototype" cluster in use now (Only tangentially related, but still interesting, is how CERN does storage in general).

How CERN does...

[Sockatume]

The trivial answer always seems to be "on an enormous scale": storage, particle collisions, construction... of course you run into problems when you start to consider whether it really does generalise to everything. Some things CERN doesn't need a huge amount of. Childcare, for instance. Or lunch. I suspect that they don't do their staff fitness weigh-in on an enormous scale, but it is quite an appealing image.

Re:

[unixisc]

Since CERN is one of the owners of Scientific Linux, would they be dependent on what RH does here: can't they just do things on their own that they need?

Comment removed

[account_deleted]

Comment removed based on user account deletion

Re:id say its the other way around.

[Nite_Hawk]

Wait, we are expected to be respectable now? That wasn't part of the deal!

The Comic

[randomErr]

I thought they were talking about the Inktank comic [inktank-studios.com].

How do you back up Ceph?

[MSG]

I saw a talk on Ceph at LISA '13 and it seemed pretty cool, except that afterward I wasn't able to find any documentation specifically about making backups of Ceph object stores.

In general, I think backup infrastructure on Linux isn't great. I'm working to make that better, but generating interest in backup infrastructure is a hard sell. Shameless plug:
https://bitbucket.org/gordonme... [bitbucket.org]

Re:How do you back up Ceph?

[chuckymonkey]

It has the ability to do RBD Snapshots, it's not a perfect solution but it does work. We're actually in the process of building a Ceph system for testing of climate modeling data. https://ceph.com/community/blo... [ceph.com]

Re:How do you back up Ceph?

[Nite_Hawk]

One of the questions when you go to large scale is how do you deal with backup in general when you have multiple PB of data. That's partially where some of the new tiering features in firefly will come in for cold storage. Having said that, we are definitely more disk friendly than tape friendly right now afaik.

Re:How do you back up Ceph?

[ScuttleMonkey]

(Inktank community guy here)

There are a number of different options for backup/disaster recovery solutions with Ceph, depending on what piece(s) of the platform you are using. For instance, the object gateways (think S3) from multiple clusters can be plugged together for multi-site replication. The CephFS and block device portions both have snapshotting built in that can be replicated offsite.

In the medium-term we're looking at having a way to replicate your entire cluster over the wire at the RADOS level (underlying object store). Longer-term we'd love to be able to offer WAN-scale replication for a single cluster and the ability to snapshot a cluster (or portions/pools therein) easily.

I hope that helps. If you have more questions hit me up on #ceph at OFTC.net IRC.

Re:

[trawg]

Great stuff. This news seems great and look forward to seeing where it goes, particularly around the open sourcing of some new tools.

We've just launched a new virtual server hosting service in .au which uses Ceph for the storage system (blatant plug: https://binarylane.com.au/ [binarylane.com.au] coming to the US soon!) We chose it after a lot of evaluation of various other systems and so far it seems to be performing really well.

The stuff on the roadmap you mentioned sounds awesome as well.

Re:

[bluefoxlucid]

Is Ceph a transactional system with an operations log for the relevant object stores, or just a haphazard backing store with some semblance of synchronization? GlusterFS needs to scan the entire filesystem after it's been offline to heal: if you offline Node 1 and place a file in /mnt/gluster/file.txt, then you must stat() /mnt/gluster/file.txt after Node 1 comes back up to propagate the file to Node 1. This provides much difficulty in auto-healing and conflict resolution.

Re:How do you back up Ceph?

[Nite_Hawk]

See this thread:

http://lists.ceph.com/pipermai... [ceph.com]