CNCF Adds Rook Storage Project for Kubernetes
The Cloud Native Computing Foundation (CNCF) announced today that it is adding a storage project that promises to greatly accelerate the number of stateful applications that could run natively on top of a Kubernetes cluster.
The Technical Oversight Committee for the CNCF has decided to accept Rook, which embeds a distributed storage system based on open source Ceph software to provide file, block and object services natively within a Kubernetes cluster. Rook was developed by Unbound and embraced by Quantum Corp. The petition to make Rook an official CNCF project was made by Mesosphere. Rook is available in alpha today, with a beta release scheduled for the first half of this year.
In total, the CNCF is now hosting 14 projects in addition to Kubernetes.
Unbound CEO Bassam Tabbara says Rook eliminates the need to rely on external storage software such as NFS to access persistent storage. Available under an open source Apache 2.0 license, the current plan is to extend Rook support beyond CEPH as the Rook project continues to evolve, Tabbara says, noting Unbound chose to support Ceph first because it’s already been battle-tested in a variety of cloud computing environments.
The Rook project is a complement to an existing Container Storage Interface (CSI) project that provided an standard application programming interface (API) for attaching external storage to any type of container cluster. CSI, like Rook, is also still in alpha.
Beyond simply providing a way to deliver persistent storage natively on top of Kubernetes, Rook should help spur broader adoption of containers and Kubernetes clusters running on bare-metal servers, Tabbara says. Today most instances of containers and Kubernetes are deployed on top of virtual machines in part to provide a mechanism to access persistent storage. The Rook project will also make it possible for IT organizations to employ standard off-the-shelf storage drives rather than relying on storage devices that have been embedded within a commercial storage system provided by a vendor that typically marks up the prices of those storage devices, he adds.
It may take some more time for the fullness of that vision for managing storage on a Kubernetes cluster to be attained. But it’s already clear to many IT professionals that clusters such as Kubernetes provides an alternative mechanism to achieve hyperconvergence in a data center by unifying compute, storage and networking using open source software. In many cases, Kubernetes increasingly will be used as an alternative approach to enable hyperconvergence that doesn’t incur or minimizes commercial licensing fees.
None of this means traditional virtual machines will disappear overnight. But reliance on bare-metal servers as an alternative to virtual machines will most certainly increase, especially in IT environments where application density is at a premium. The number of containers that can run on a bare-metal server compared to a virtual machine is several orders of magnitude higher.
There are also efforts underway to more closely align containers with a lighter-weight hypervisor. That approach would provide all the isolation enabled by a virtual machine without incurring additional overhead that is either not needed or duplicates functionality already embedded inside the container cluster.
Obviously, both Rook and CSO have a long way to go before they fully mature. But the one thing that is clear is that the management of IT infrastructure is slowly but surely about to be utterly transformed once again.