Red Hat Ceph Storage 3 brings file, iSCSI and container storage

Open source software distributor Red Hat has released version 3 of its Red Hat Ceph Storage software-defined storage.

This adds file access storage, iSCSI block storage and storage deployed through Linux containers to the software-defined storage product, mostly aimed at OpenStack deployments.

The CephFS file system already existed as a “tech preview”, said Gerald Sternagl, business unit manager for Red Hat. This allowed customers to test it, and it has now been released for production workloads.

Sternagl said CephFs has been built from scratch and is a Distributed File System. It is based on the use of discrete metadata servers that help coordinate data flows for better parallel operations.

“CephFS can scale across the whole datacentre. It’s easy to add more Ceph storage nodes to get unlimited scalability, potentially into petabytes and beyond,” said Sternagl. 

Use cases envisaged are mostly around OpenStack, through Manila, but Red Hat may provide NFS or SMB access in future, he added.

Ceph’s developer, Inktank, was acquired by Red Hat in April 2014. The software remains open source and is a software-only product based on multiple storage nodes and a technology called Rados (reliable autonomic distributed object store) that lays out and manages data across multiple clusters.

Meanwhile, Red Hat has taken open source iSCSI software and written it into Red Hat Storage to be redundant across multiple nodes.

Ceph already provided block storage through Rados Block Device (RBD), mostly as a means of providing block for OpenStack.

Key use cases for iSCSI are to provide protocol connectivity for existing scenarios, in particular VMware deployments, said Sternagl.

The new version of Ceph also provides storage in containers. Previously, Ceph storage deployments required a minimum of seven hardware nodes, on which different components of the software had to be deployed.

This can now be done in three hardware nodes because the various software components – such as management layer, object storage daemons and metadata servers – can be bundled into containers that can be co-located, whereas previously that wasn’t possible. The key advantage is a hardware saving in a ratio of 3:7.