buchachon - Fotolia
Red Hat has recast its storage focus from stand-alone Ceph and Gluster products to OpenShift, OpenStack and hyper-converged technologies that can target enterprise trends in DevOps, hybrid cloud and edge computing.
At least parts of the revised Red Hat storage strategy seem to be finding a receptive audience. For instance, sessions on the company's OpenShift Container Storage (OCS) were packed at the recent annual customer conference, and the product management director claimed the software has more than 400 users with data stores up to 300 TB.
"Customers don't come to us and say, 'Give me storage,'" said Ranga Rangachari, vice president and general manager of Red Hat storage. "They come to us and say, 'I've deployed OpenStack. Give me the best storage for OpenStack.' Similarly, they go and deploy OpenShift and say, 'Give me the best solution for OpenShift storage.'"
Red Hat's storage strategy initially focused on the Gluster file system, acquired in 2011, and multiprotocol Ceph software, which was added in 2014. Like all of Red Hat's open source software, the storage products run on commodity server hardware. The company pitched customers on the potential cost savings in hopes that its software-defined storage would become as popular as its enterprise Linux distribution and challenge traditional storage systems that tie software to proprietary hardware.
Progress is tough to gauge, because Red Hat doesn't break out its storage revenue and declines requests for storage customer counts. But the company is finally starting to share limited details as it shifts to a hybrid-cloud-centric strategy.
New Red Hat storage pillars
The three pillars of Red Hat storage are now OpenShift Container Storage, hybrid cloud exascale object storage and hyper-convergence. Rangachari said OCS has the most customers, and the hyper-converged and object storage products have more data stored, thanks to large telco deployments.
Gluster plays a smaller role than Ceph in the new three-pillar focus. Only the virtualization option of the Red Hat Hyperconverged Infrastructure product uses Gluster. Red Hat's Hyperconverged Infrastructure for Cloud and hybrid cloud object storage are built on Ceph, and OCS is moving from Gluster to Ceph.
The upcoming 4.x release of Red Hat's OpenShift Container Storage will change to Ceph, providing a more complete platform of block, file and highly scalable S3-based object storage. Rangachari said the move from stateless to stateful container-based applications, inherently with more data, is driving the need for more scalable Ceph object storage. OCS provides persistent storage for Red Hat's separately sold OpenShift container application platform.
Many Red Hat customers currently use third-party storage with OpenShift deployments, but plenty were checking out OCS at the annual customer conference. Jason Tower, a senior infrastructure engineer at Blue Cross Blue Shield of North Carolina, said Red Hat's OCS could give the IT team a chance to take advantage of underutilized storage drives in its OpenShift environment.
Blue Cross Blue Shield N.C. currently uses dedicated hardware from NetApp and Dell EMC to provide persistent storage for its container-based applications. Tower said the company's OpenShift hardware tends to be 1U servers or blades that are light on disk storage, so the infrastructure team found it more efficient to go with the existing external storage.
"Like everyone else, I'm being asked to do more with less," Tower said. "I certainly want to keep my options open. I'd just as soon not be beholden to other vendors and avoid any possibilities of lock-in."
Royal Bank of Canada is checking out a variety of Container Storage Interface-compliance storage options, including Portworx and Red Hat's OCS, according to Shahtab Khandakar, an associate director at the firm. He said Royal Bank is a big NetApp shop and initially went with the "natural choice," deploying the Trident plugin to connect its NetApp storage into its OpenShift environment.
"Now, things are changing so rapidly, this [OCS] might be another option," Khandakar said.
Companies that are starting to develop container-based applications are exploring Red Hat storage. Dallas-based City Electric Supply is "very legacy-oriented" with traditional SAN technology, according to Chris Lanteigne, head of infrastructure and operations. But Lanteigne said he is looking into Red Hat's OCS with Ceph because the company is shifting to a DevOps model.
Ceph with OpenStack
Ceph's roots are even deeper in OpenStack cloud deployments, where it has been a popular back-end storage choice for years. Rangachari said the first wave of Ceph customers typically deployed OpenStack with block-based storage, storing no more than a couple of petabytes. He said the new wave of customers tends to use Ceph object storage to scale to tens if not 100 or more petabytes. The range of users extends from managed service providers selling public cloud storage services to organizations building a private cloud infrastructure, Rangachari said.
Symcor, a Canadian provider of business and payments processing services, plans to roll out Ceph in production within the next year as it shifts from traditional big iron, Fibre Channel SAN and virtualization technology to an OpenStack private cloud, according to Greg Procunier, senior architect at Symcor. He said Symcor has used Ceph for block and S3-based object storage in the lab for the past few years, and the object store could work well for use cases such as check archival.
"We're looking to modernize our data center infrastructure and move to a more agile methodology of deploying compute and applications. So, the move to the cloud was inevitable," Procunier said. "Due to compliance requirements, we're not able to move into the public cloud space."
Procunier said he likes the fact that Ceph supports block, file and object storage. He cited CERN as an example of "an extremely performant Ceph deployment" that gives him confidence to use the technology in production.
"It's not just about cheaper or faster," Procunier said. "It's about smarter ways of getting at the data. There's a lot of really interesting possibilities with this technology."