This is a guest post by Kirill Shoikhet is CTO of Excelero Storage
Enterprise storage caters to a wide variety of application stacks, and as digital transformation efforts add new platforms and paradigms to the infrastructure, maintaining storage flexibility is critical. As a result, storage architects are keen to deploy storage that is invisible to their applications. Best practice in traditional environments has been either to use storage appliances or kernel-based Software-Defined Storage. Both present virtual “local” drives to the applications.
In container environments, the same invisibility is easily achievable using a CSI plug-in with “container-ready” storage where storage runs alongside application containers. To date, most traditional storage vendors have chosen this approach.
Recently, several “container-native” vendors began to offer solutions that enable organizations to build a storage infrastructure that runs inside containers and also is controlled by containers. Access occurs via the same CSI plug-in, but does not require a separate storage entity, be it an appliance or a storage cluster.
On-Premise and In the Cloud
As Kubernetes and other containerization orchestration platforms are incorporated into mainstream infrastructures, a growing community of developers has needed to deploy K8s at scale, both with on-premises workloads and in the cloud for applications with often widely differing needs. That is why over 2,000 organizations trust OpenShift as the industry’s leading enterprise Kubernetes platform that enables native support for hybrid and multicloud deployments.
However, for these developers, storage often has been a challenge. Even those architectures considered “container native” can lose a performance edge, since they put some or all of their data path into the containers, which can make it hard to provide predictable low-latency storage access. For many enterprise applications, this can be a deal-breaker.
To date, storage options for Kubernetes have included:
- On-premises and co-located appliances/storage clusters. While on-premise storage is often the most feature-rich option, it can often be challenging in cloud- and container-native deployments with impact to digital transformation initiatives. Even newer NVMe- and kernel-based enterprise software-defined storage can require significant rewriting to become container native.
- Container-only storage software. Solutions that were born as containers and implemented using them. These products take the “functionality-first” approach and address performance, and performance-at-scale and in production , as a follow-on consideration. Published data from vendors falls well below that of bare-metal high-performance storage options, even with results framed from best-case scenarios.
- Software-Defined Storage running natively in Kubernetes. A new option for the Red Hat community when using Excelero NVMesh alongside Red Hat OpenShift that has been shown to benefit IT flexibility, scalability, and costs.
How Excelero NVMesh Works With OpenShift
Deploying both OpenShift and Excelero’s NVMesh software-defined storage is an option when IT teams are looking for enterprise-grade container-native solutions with predictable performance, regardless of where the storage is deployed, either on-premises or on public clouds, or a combination of both.
With Excelero NVMesh’s native support for RHEL CoreOS through its NVMesh Operator acting as a custom controller to install and control NVMesh inside OpenShift clusters, IT teams can build agile storage infrastructures without the need for an external storage cluster or appliance. They can support public cloud, hybrid cloud, multicloud, and on-premise models with the same technology stack with the customer’s choice of RDMA and TCP/IP based storage networking. Backed by Excelero NVMesh, datacenters teams using OpenShift can create agile and elastic infrastructures that achieve superior storage performance, scalability, and ROI, with the flexibility to adjust further as their needs grow.
For example, a CI/CD environment with cloud bursting replacing bare metal servers, or a large number of single node databases like MariaDB deployed within containers. In both cases, application customers expect performance comparable with bare metal servers with NVMe drives while infrastructure management teams expect the scalability, agility, and cost efficiency of Kubernetes.
In testing on Microsoft Azure HBv3 virtual machines, Excelero NVMesh transformed available Azure compute resources into a storage layer on par with typical on-premises High Performance Computing (HPC) cluster storage. With testing as wide as 24 nodes and 2,880 cores, latencies, bandwidth, and IO/s levels scale well. As the number of nodes consuming IO was increased, the system performance remained consistent. The results also show linear scalability: As additional target nodes are added, the system performance remains consistent. In subsequent testing, Excelero replicated the same level of performance under a RH OCP cluster running on top of the same Hbv3 VMs.
For the same reasons, a large European government body agency dealing with cyber initiatives and demonstration of new IT technologies has also leveraged Excelero NVMesh with OpenShift for performance-critical tasks.
Data Protection and Container-Native Storage
When combined with Excelero NVMesh’s scalable distributed erasure coding (EC) for data protection, datacenters can obtain unparalleled high throughput and low latency with even a minimal configuration. NVMesh’s performance scales linearly with cluster size, and its distributed EC provides dual-protection from drive and node failures. The performance and especially the performance density that can be achieved with such building blocks at datacenter scale can propel the cost and time savings that are critical to modern architectures.
IT leaders are rethinking the standard requirement to deploy high-performance storage appliances with Container Storage Interfaces (CSI) plug-ins. The dependency on appliances can restrict hybrid and multicloud deployments of these workloads and slow down digital transformation.
When storage architects integrate Excelero’s NVMesh software-defined storage and OpenShift, they can enable fluidity in hybrid deployment of containerized applications backed by the familiar OpenShift experience and Excelero NVMesh’s scalable, high-performance, agile storage.
Original article is featured at cloud.redhat.com.
As a leading kubernetes storage provider, Excelero offers the performance of local flash with the persistence and scalability of centralized storage. Improve oracle database performance and achieve high throughput and low latency for mission-critical workloads with Excelero’s oracle cloud database solutions.