LINBIT today announced it has made support for Kubernetes clusters generally available on its open source software-defined storage (SDS) software that now integrates with the open source MariaDB relational database.
The LINBIT SDS offering combines open source LINSTOR volume and configuration management software for block storage and DRBD, the robust in-kernel data replication software, with open source Distributed Replicated Block Device (BRBD) software that is employed widely at the Linux kernel level to replicate data to enable disaster recovery and high availability.
LINBIT COO Brian Hellman says LINBIT SDS provides hosts running Kubernetes with access to persistent block storage for Kubernetes environments, including support for logical volume management (LVM) snapshots, thin provisioning and configuration management.
The rate at which stateful applications are being deployed on Kubernetes is starting to accelerate, says Hellman. As those applications move into a production environment, they require access to block storage that is persistent. Rather than having to acquire a separate storage array, LINBIT makes it possible for hosts running Kubernetes to access storage directly attached to the server.
Hellman says one of the primary benefits of the LINBIT approach is the minimal amount of overhead that gets introduced because it operates at the kernel level. As servers based on significantly faster NVMe interfaces become more widely employed in the data center, organizations should see as much as 30 percent improvement on top of that, he notes.
It’s still early days in terms of deploying Kubernetes in production environments. Relational databases usually wind up being deployed on virtual machine and physical servers in equal measure, depending on the performance requirements of the application running on top of them. Access to block storage on the server should encourage more organizations to consider deploying Kubernetes on bare metal servers to reduce the overall size of the software stack deployed on a physical machine. That approach not only should improve application performance, but also it may reduce overall costs by eliminating the need to license commercial virtual machine software.
LINBIT, of course, is not the only IT vendor working on approaches to making persistent storage available to containerized applications and Hellman concedes there is some overlap in those efforts. But as storage and I/O performance issues in general receive more attention, the rate at which Kubernetes gets deployed in production environments should increase substantially in the months ahead. The issue many organizations will need to address is whether they want their storage resources managed by the team that manages the Kubernetes clusters or by a dedicated storage administrator. In many cases, the adoption of Kubernetes clusters will force a larger DevOps discussion concerning roles and responsibilities in a modern enterprise.
In the meantime, IT organizations should expect to see more databases showing up within a container as a new generation of microservices-based stateful applications inevitably make their way into production environments. Chances are, given the latency sensitivity of many of those applications, the only real option is going to be to attach as much storage as possible to the Kubernetes host.