gjp311 - stock.adobe.com
Red Hat rivals VMware and Nutanix with hyper-converged OCS 4.5
Red Hat’s OpenShift 4.5 extended container orchestration to virtual machines. The arrival of OCS 4.5 moves the solution into the realm of hyper-converged systems
Red Hat’s offer around OpenShift has never come as close to those of Nutanix and VMware as it does now.
With the release of OCS 4.5, it can present virtual machines, containers and storage, and allow them to be administered from the same interface in hyper-converged infrastructure fashion.
OCS is OpenShift Container Platform, which is an implementation of OpenShift with Ceph storage as its back end. New OCS 4.5 functionality adds to that in July’s version 4.5 of Red Hat’s OpenShift Kubernetes distribution, and allows it to rival vSphere 7 Tanzu and AOS from Nutanix with the Karbon extension.
The software-defined OCS 4.5 acts like a hyper-converged infrastructure solution, with management possible via the same interface used to administer external storage. That’s like VMware’s vSAN, which now interfaces with containers from its new vSAN Data Persistence module, and like Nutanix, which works with the same types of storage as Ceph.
“Historically, OCS allowed OpenShift to manage local storage and to attach it to containers it ran,” said Yacine Kheddache, architecture chief with Red Hat France.
“But since the arrival of OpenShift 4 and the shift towards Kubernetes our customers have wanted to attach to external storage from OpenShift, which has been possible from version 4.2 of OCS, thanks to the replacement of its Gluster engine with Ceph.
“The new version 4.5 of OCS finalises this integration with a dedicated graphical admin tab for OpenShift and the possibility of triggering automated processes,” he said.
In its standalone version, Ceph is software-defined storage installed on a cluster of servers with multiple drives on which storage volumes are created, in block, file or object mode.
Its key advantages lies in performance and its great elasticity. Its key weakness, however, is that it can be very complex to manage from its own interface.
OpenShift, meanwhile, radically simplifies containers orchestrated by Kubernetes. Indeed, it contains all the interfaces to allow developers to deploy resources without having to call on system administrators.
“The first enterprises interested in integration of Ceph into OCS are those that use OpenShift and Ceph storage because it permits them to manage Ceph with all the facility offered in OpenShift without having to jump from one interface to the other,” said Kheddache.
However, he said, use cases encountered in a Ceph cluster often go beyond those of a joint workload with OpenShift. In that case, volumes that aren’t used with containers but which support applications run by other servers are administered from Ceph’s own interface.
But the guess would be that Red Hat’s aim is that external servers should be also orchestrated by OpenShift since it’s also capable of running virtual machines.
OCS can administer volumes on disks internal to OpenShift nodes and on those of an external Ceph cluster. It can’t deploy volumes that straddle the two but instead offers a mechanism for replication or migration between the two.
In hybrid cloud mode, the OCS console deployed in OpenShift on-site can also manage OCS storage volumes on OpenShift instances in AWS and handle automated migrations between the two.
“In AWS, OCS volumes use EBS [Elastic Block Store] services,” said Kheddache. “We don’t officially support any service except this service for now, because to perfect integration we have developed management with the functionality extended by the AWS API.”
“Nevertheless, version 4.5 of OCS comes with preview modules that provide compatibility with Blob storage, which is the equivalent to EBS in Azure.”
Version 4.5 of OCS and OpenShift bring a number of other new features too. One of these is the ability to deploy in “bunker” mode, in which no additional packages need to be downloaded during installation. This functionality has been developed to respond to the needs of customers with sensitive activities and that want a gap between their datacentre and the internet.
“Installation packages are made available on local servers,” said Kheddache. “And in numerous cases, our customers feed the server with updates they bring on removable disks.”
OCS 4.5 also comes with the new end-to-end encryption protocol FIPS-140.
More on hyper-converged infrastructure
- SAN vs NAS vs hyper-converged for virtual machine storage. The pros and cons of file and block access storage arrays vs the new breed of all-in-one server/storage architecture for storage of virtual machine systems and data.
- 10 top hyper-converged infrastructure vendors and products. Determining which hyper-converged infrastructure platform is right for your organisation should include analysing your workloads and assessing technology from top HCI vendors.
Meanwhile, OCS 4.5 now supports two servers being able to write to a block storage volume at the same time. This function is intended to accompany the arrival of virtual machines in OpenShift 4.5 in allowing hot migration of a VM between clusters, which is equivalent to VMware’s vMotion or Nutanix’s Live Migration.
“It’s interesting to note that we have never needed to carry out hot transfers of applications between two OpenShift clusters,” said Kheddache “Because good practice with containers is to always re-launch a new copy of the instance. But, we’ve shown that we are capable of adapting very rapidly.”
OCS also revives functionality that Ceph has forgotten: Accessing block volumes without having to go through the S3 protocol.
“Ceph was originally conceived with access via RADOS and it’s using this that we have installed an S3 compatibility layer,” he said.
“To utilise RADOS directly can nevertheless prove useful in the case of applications that don’t have the ability to be exported to the cloud, because access is very fast. With the low level protocol available in OCS you can install it on top of any other object protocol,” said Kheddache, with the possibility to migrate applications written for older object storage systems in mind.