Systems Engineer
Known by many as providing a commercial open-source Linux distribution, Red Hat has somewhat quietly been a player in storage for some time now. Many users of enterprise storage may not even be aware, but Red Hat has been involved in several projects for Software Defined Storage (SDS). Most prominently, they have been contributing to Ceph and the Gluster filesystem, both of which have Red Hat Subscription based products of the same names. Moreover, Ceph has been at the core of storage presented in OpenStack with Swift, for businesses building private clouds. It was also the basis of Red Hat Container Native Storage, which was later renamed as OpenShift Data Foundation (ODF). Recently, Red Hat and IBM made some big news by announcing that the ODF/Ceph development team and products will be moving under IBM’s storage business.
As you can tell by the names “Container Native Storage” and “OpenShift Data Foundation”, Red Hat put considerable focus on making Ceph the center of persistent storage for the OpenShift platform, and it works well for that.
What is Ceph?
If you’re new to Ceph, I’ll give a brief description. Ceph is based on a type of cluster called “RADOS” for Reliable Autonomous Distributed Object Store. It is typically built with Storage Rich Servers (the type of server that you can load up with HDDs), and is able to manage data protection (via Erasure Coding or Replication, depending on the needs of the data and the application). Access to the data can be through LIBRADOS, which is a library that allows applications written in several popular languages to access the storage directly. It is also possible to use the RADOS Gateway (RADOSGW) for S3 or Swift compatibility, Ceph FS for NFS access, or RADOS Block Device (RBD), which allows Ceph to be used as block storage with iSCSI.
In a Red Hat OpenShift environment, where “Read-Write-Many” (RWX) access is critical for most applications, Ceph is still a natural fit for providing persistent storage to containers via NFS. Alternatively, there is Read-Write-Once (RWO), which would be the option for typical block storage. The limitation to this is that you cannot have a pod (a group of containers in an application) share a RWO device. So, if you’re scaling out an application using OpenShift Container Platform (OCP), it really is critical to have RWX infrastructure for most of your applications as well as the internal functions of the cluster.
So, what’s the problem?
The problem is that Ceph/ODF are not the “core” of the technology here. OpenShift (Kubernetes, Red Hat CoreOS, and the other capabilities that are packaged together) is the core product. There are some very large users of Ceph, typically in educational and research institutions. And Ceph is really good at allowing commodity hardware to provide extremely large-scale storage at a low cost, if you have the personnel (or Grad Students) to manage the hardware. But it never seemed to really hit with the market. The main push that Red Hat has made with Ceph has been as part of ODF, but that is more an enabler to help customers adopt Red Hat OpenShift than a core part of its portfolio.
The fact of the matter is that there are many storage providers that will work with OpenShift. Red Hat partners with Nutanix to provide a Hyper-Converged Infrastructure (HCI) solution. IBM offers Spectrum-Scale (formerly known as General Parallel File System or GPFS) and IBM Spectrum Fusion, including the Spectrum Fusion HCI platform. NetApp has Trident. The list goes on, and I apologize to any vendors that I left out. The fact is that Red Hat has a strong interest in being neutral and just staying focused on developing OCP to work with any Container Storage Interface (CSI) compliant vendor.
So, what to do? One could just give up and allow the existing customers to go End of Support, but that could be a tough pill for them to swallow. Additionally, there is a vibrant Ceph Open-Source community (the most recent release is Quincy – each Ceph release is named after a type of Cephalopod), and having involvement in that community is beneficial for everyone involved.
In contrast to Red Hat, IBM’s Storage business is one of the core pieces of its Systems Group. Just this week, Gartner again placed IBM Storage in its “Leaders” quadrant on the Magic Quadrant for “Primary Storage”. IBM has, in fact, been selling Red Hat OpenShift Data Foundation (offering a free version for up to 12 TB) as part of its IBM Storage Suite for IBM Cloud Paks. Beyond that, IBM has been offering IBM Spectrum Fusion (see mySpectrum Fusion HCI blog), as a means of providing Cloud Native storage for Red Hat OCP.
So, IBM is very focused on providing the resources to drive storage business in the container native world. But, Red Hat is a separate company, so why would IBM turn over its IP, let alone its development team? Well, there’s precedence for this.
Red Hat Advanced Cluster Management (ACM) had its roots in IBM and was a technology shared to Red Hat where the benefits would be strongest. Also, more recently Red Hat’s Process Automation Manager and Decision Manager began the process of transitioning to IBM. This latest exchange, with the entire Ceph Development team moving to IBM, absolutely has echoes of this.
What does this mean for Ceph?
Well, in the short term it means that Ceph has a future. It’s not just going to be sunset/End of Service Lifed. It’s going to be developed moving forward and will be part of the larger IBM storage portfolio. It means that IBM is dead serious about taking as much as they can in the Container Native Storage space. They have resources to pour into the Upstream project (which will continue to be an Open-Source project) and resources that are dedicated to selling storage as their focus area.
In the announcement material, IBM stated that Ceph/ODF will become the foundation of IBM Spectrum Fusion (which currently uses Spectrum Scale Erasure-Code Edition). They, in fact, also intend to use Ceph with Spectrum Scale to provide a “data lakehouse”.
In there here and now, though, it means that if you’re looking into deploying OCP, IBM is prepared to give you options.
More Information
For more information on best storage solutions for your apps and workloads or to gain more efficiencies from your current data storage environment, contact me or your Mainline Account Executive, or reach out to us here with any questions.
You may be interested in:
BLOG: IBM Spectrum Fusion HCI – Hybrid Cloud Infrastructure for the Enterprise