Deployments
Simplyblock is a highly flexible storage solution.
Different initiator (host) drivers (Kubernetes CSI, Proxmox, OpenStack) are available. The storage cluster deployment can be installed into Kubernetes (disaggregated or hyper-converged) or via Docker (also called "Plain Linux" deployment). The Docker-based deployment is fully deployed and managed via the Simplyblock CLI or API, minimal Docker knowledge is required.
Control Plane Installation
Each storage cluster requires a control plane to run. Multiple storage clusters may be connected to a single control plane. The deployment of the control plane must happen before a storage cluster deployment. The control plane can be installed into a Kubernetes Cluster or on Plain Linux VMs (using Docker internally). For details, see the Control Plane Deployment on VM or Install Control Plane on Kubernetes
Storage Node Installation
For details on how to install the storage cluster into Plain Linux, see Install Simplyblock Storage Nodes on Linux.
For installation of Storage Nodes into Kubernetes, see here: Install Storage Nodes on Kubernetes
Installation of Drivers
Simplyblock logical volumes are NVMe over TCP or RDMA (ROCEv2) volumes.
They are attached to the Linux kernel via the provided nvme-tcp or nvme-rdma
modules and managed via the nvme-cli tool. For more information, see
Linux NVMe-oF Attach.
On top of the NVMe-oF devices, which show up as linux block devices such as /dev/nvme1n1,
life cycle automation is performed by the orchestrator-specific Simplyblock drivers:
- On Kubernetes: Simplyblock CSI Driver
- On Proxmox: Proxmox Integration
- On OpenStack: Cinder Driver
Generally, before creating volumes it is important to understand the difference btw. an NVMe-oF Subsystem and a Namespace.
System Requirements and Sizing
Simplyblock is designed for high-performance storage operations. Therefore, it has specific system requirements that must be met. The following sections describe the system and node sizing requirements.
For deployments on hyper-scalers, like Amazon AWS and Google GCP, there are instance type recommendations. While other instance types may work, it is highly recommended to use the instance type recommendations.