Red Hat bolstered Gluster storage for its OpenShift Container Platform, adding iSCSI block and S3 object interfaces,...
as well as greater persistent volume density.
Red Hat Container-Native Storage (CNS), built on Red Hat Gluster Storage 3.3, provides storage for OpenShift developers. The CNS 3.6 release is designed specifically for the 3.6 version of the OpenShift Kubernetes container application platform Red Hat launched in August.
Red Hat added an iSCSI block interface to the file-based Gluster Storage to enable its use in container environments with distributed databases and other low-latency workloads, such as Elasticsearch. The company also made available a technology preview of a container-based object store inside OpenShift for applications that require an Amazon Simple Storage Service (S3) API.
The latest Red Hat CNS release enables customers to store not only application data, but also core infrastructure elements, such as OpenShift registry, logging, and metrics data. The 3.6 release also supports three times more persistent storage volumes than the product did in the past, translating to about 1,000 persistent volumes per cluster, according to Irshad Raihan, a senior manager of product marketing at Red Hat.
'Complete storage platform' for OpenShift
Sayan Saha, director of storage product management for Gluster at Red Hat, said CNS 3.6 provides "a complete storage platform" for OpenShift users. He said the impetus for CNS 3.6 was not "to go after regular block or object use cases."
Saha said the CNS addition of low-latency block storage would enable the level of performance required by workloads such as Elasticsearch for logging and Cassandra for metrics in the OpenShift infrastructure. Most OpenShift users relied on local storage to address those needs in the past, he said.
"Customers would come to us and say, 'OK, you've got persistent storage for OpenShift with Gluster, but what about my [Elasticsearch] and Cassandra?'" Saha said.
He said OpenShift users often have multiple types of applications that require different access modes with Kubernetes. Kubernetes enables container-based applications to access persistent storage volumes in three different ways: ReadWriteOnce (RWO), ReadOnlyMany (ROX), and ReadWriteMany (RWX). RWO enables a storage volume to be mounted by a single node or container, while ROX and RWX enable a volume to be accessed by multiple nodes or containers.
Saha said RWX is fundamentally a shared file use case that Red Hat enabled through its Gluster FUSE client. He said the newly added iSCSI interface would apply to RWO block storage uses.
Kubernetes OpenShift does not support an object interface by itself, according to Saha. But he said Red Hat implemented object storage inside OpenShift to enable application developers with cloud-native applications to use S3 buckets with Gluster back-end storage if they need it. He said many developers coming out of college only know the S3 API, "so if you don't have that capability, it becomes an incomplete story."
Saha said Gluster storage always supported the S3 API through OpenStack Object Storage's Swift API, but Red Hat did not emphasize it because Gluster users focused on the file interfaces.
Red Hat supports the S3 and iSCSI interfaces only for container use cases with a Gluster storage back end, Saha said. He said a customer would not be able to use the iSCSI interface for stand-alone block storage with VMware or use the S3 API with Amazon S3 public cloud storage.
Sayan Sahadirector of storage product management for Gluster, Red Hat
Saha said Red Hat uses Gluster Storage, as opposed to its Ceph storage product, with OpenShift because Gluster can run everywhere -- in on-premises and public cloud environments. "OpenShift with CNS can actually provide a consistent experience for both compute and storage across hybrid and multicloud environments," he said.
Pricing for Red Hat Container-Native Storage is per node. A three-node CNS premium configuration costs $12,000, and a three-node standard configuration is $9,000.
Henry Baltazar, a storage research director at 451 Research, said making block and object interfaces available are "not trivial additions" for containers. "It takes many years for storage vendors to add protocols like that," he said.
Baltazar said many vendors are adding container storage because of the level of development taking place. He said when developers select storage, they may go with other types of vendors they already know, rather than traditional storage vendors. And that will help Red Hat.
"You can't really develop large-scale persistent applications without storage. So, sooner or later, they're eventually going to need storage functionality. And when they go shopping for storage, the easiest thing for them to do would be something they're already familiar with. For many, that would be Red Hat," Baltazar said.
Red Hat storage roadmaps focus on containers, HCI
Deep dive into Red Hat OpenShift
Red Hat users try to adapt to container paradigm shift
Weighing GlusterFS vs. Ceph