Hyperconverged architecture

In the context of hyperconverged nodes, a minimum of six nodes is required to be eligible for Ubuntu Pro Managed Solutions and Delivery services with Canonical.

We’ll provide three reference specifications based on the purpose of the storage environment, i.e. performance, capacity and general-purpose storage.

Performance

In the trade-off between cost and performance, performance-oriented nodes prioritise low latency, high throughput, and high concurrency.

Processor	56 cores (x86) or greater
Memory	256 GB RAM
Network	2x dual-port NICs (25 Gb+), onboard NIC for OOB
Storage	2x 960 GB SSD/NVMe for OS; 8x enterprise NVMe for OSDs

Capacity

Available storage disk capacity (what users will see).

Processor	32 cores (x86)
Memory	256 GB RAM
Network	2x dual-port NICs (25 Gb+), onboard NIC for OOB
Storage	2x 960 GB SSD/NVMe for OS; 4x enterprise SSD/NVMe for RadosGW metadata OSDs (4% NL-SAS capacity) WAL/DB*; 20x NL-SAS or SSD for OSDs

Note

Separate WAL/DB or bcache disks should only be purchased if there is a significant difference in performance between the OSD disks (slow disks, i.e. “Very Read Intensive SSDs”) and the dedicated WAL/DB disks (fast disks).

General-purpose storage

A balanced approach between available capacity and performance.

Processor	64 cores (x86) or greater
Memory	256 GB RAM
Network	2x dual-port NICs (25 Gb+), onboard NIC for OOB
Storage	2x 960 GB SSD/NVMe for OS; 20x enterprise SSD/NVMe for OSDs