Hardware Recommendations
Hardware Recommendations
The euroNAS HA Cluster is not tied to specific hardware and runs on standard x86 server platforms, giving full flexibility for initial deployment and future expansion.
CPU
For optimal performance, server-grade processors from Intel or AMD are recommended, such as Intel Xeon or AMD EPYC. Dual-socket configurations can be used where additional compute capacity is required.
Installation Drive
The operating system should be installed on a dedicated SSD, with a minimum capacity of 250 GB. This drive is reserved for the system only and is kept separate from data storage. It is not intended for hosting application data or storage workloads.
Storage Drives
A separate drive or pool is required for data storage. This can consist of individual disks or RAID configurations, depending on performance and resilience requirements. Server-grade drives are strongly recommended, as consumer-grade disks are not designed for sustained workload or failure scenarios. Hybrid configurations are supported, allowing slower SATA or SAS drives to be combined with SSD or NVMe devices for caching and performance optimisation.
System Memory
A minimum of 16 GB RAM is required, with additional memory recommended depending on workload and storage configuration.
Network
A minimum of two network interfaces is required, although additional interfaces are recommended for resilience and performance.
The cluster separates heartbeat traffic for node health and coordination, replication traffic for data synchronisation between nodes, and client access traffic for external connectivity.
For best practice, use dedicated interfaces for replication traffic and avoid mixing replication and client traffic where possible. 10 GbE is recommended for replication to ensure consistent performance. Direct node-to-node connections can be used for replication where appropriate, although switched configurations are also supported.
For higher availability, use multiple interfaces with bonding for resilience and throughput. Avoid placing all bonded links on a single physical NIC, and where dual-port cards are used, ports should be split across bonds to avoid a single point of failure.
Static IP addressing should be used. DHCP is not recommended, as IP changes can disrupt cluster communication.