Architecture

Edit online

IBM Storage Ceph has a high availability scale-out software-defined software (SDS) architecture. The IBM Storage Ready Nodes are interconnected with high-speed data switches that provide both client access and resilient storage for their data.

The following architectural diagram presents a general overview of the physical components that are involved in an IBM Storage Ready Node platform:

Figure 1. Architectural diagram
Architectural diagram

IBM Storage Ready Nodes supports IBM Storage Ceph SDS. The use of two data switches facilitates both high availability (HA) and higher interconnect bandwidth through link aggregation under normal operating conditions. Data switch uplinks connect to the facility network in HA spine-leaf configuration. Client connections are made by using facility connections to the data switches. The separate Out of Band (OOB) management switch network facilitates IBM Storage Ready Nodes hardware and firmware platform management. Each Ready Node has a separate network connection to the management switch and a related console.

The following sample rack stack-up image depicts a dedicated Ready Node and switch rack configuration:
Figure 2. Sample rack stack-up
Sample rack stack-up
Note: The rack, PDUs, and the associated switches are customer-supplied. For more information on specifications, see Site preparation and pre-installation. The customer orders the IBM Storage Ready Nodes and the product is delivered to the customer site for installation.

The whole system of IBM Storage Ready Nodes has the following characteristics:

  • A data center rack capable of supporting IBM Storage Ready Nodes and network switches.

  • Rack PDUs with sufficient capacity and 1+1 high availability. These PDUs power rack components within local electrical code requirements.

  • Multiple Ready Node servers that support customer capacity, performance, and data availability and integrity needs.

  • Data and management switches that provide interconnections between the Ready Nodes, the data center network, and management console.

  • System-level high availability (HA) capabilities include:
    • 1+1 HA PDU line power to all the components.
    • 1+1 HA sharing power supply units (PSU) integrated into Ready Nodes and switches.
    • 1+1 HA data networking with link aggregated data network connections for highest throughput.
    • Ready Node mirrored OS drives.
    • IBM Storage Ceph-native data availability and resilience capabilities.