Hardware

Choosing the right hardware is an important step in planning your Gallium Cluster. This page covers system component requirements and Gallium's hardware support policies.

System Components

System Requirements

Each server in a Gallium Cluster must meet the minimum requirements listed below. The Gallium Console will prevent installation on servers that do not meet these specifications.

Resource	Minimum	Recommended
Memory	32 GB RAM	64 GB RAM
CPU	8 cores / 16 threads	16 cores or more
Data Disks	SSD	NVMe
Network Adaptors	2 × 10 Gbps + 2x 1 Gbps	4 × 10 Gbps

All hypervisors running the Worker role must have identical CPU and memory configurations. This ensures virtual machines can be live migrated and failed over between any Worker hypervisor in the cluster.

System Disk

Each server requires a dedicated System Disk for the Gallium installation. Supported options include:

• A single SSD — 240 GB minimum, 480 GB recommended.
• A boot-optimised storage device such as the Dell BOSS, HPE NS204i-u or Cisco UCS-M2-HWRAID.

When using a single SSD redundancy is provided by total failover to another hypervisor within the cluster.

Cluster metrics are written to the system disk on each hypervisor. This can generate ~100GB per day of write activity on busy clusters. Ensure your system disk is appropriately rated for a DWPD of 0.5 or higher.

Storage Adaptors

HBA (Host Bus Adapter) controllers are preferred for cluster deployments. RAID controllers are not recommended — Gallium manages disk redundancy at the software level through virtual disk copies, so hardware RAID provides no benefit and can interfere with storage operations.

RAID controllers placed in passthrough mode may still limit the hypervisor's ability to monitor disk health. When in doubt, use purpose-built HBAs or controllers placed in IT mode.

Network Adaptors

Use server-grade 10 Gbps (or better) network adaptors. Intel NICs are recommended for best compatibility.

Thunderbolt and USB network adaptors are excluded.

NIC Feature Requirements

Network adaptors must support the following settings.

Setting	Required Minimum
Transmit Queues	1
TX Ring Size	4096
Receive Queues	8
RX Ring Size	4096
Completion Queues	9
Interrupts	11
Receive Side Scaling (RSS)	Enabled

Most server-grade Intel NICs support these settings out of the box. Consumer or desktop-class adaptors may not support the required ring sizes or queue counts.

Network Switches

Industry standard managed switches are sufficient. There is no requirement for special switch functionality.

For redundancy, deploy an A-side and B-side switch. Each server's network adaptors should be distributed across both switches so that a single switch failure does not take the cluster offline.

The link between the A and B switches should be sized at the infrastructure NIC speed multiplied by N-1, where N is the number of servers. This ensures sufficient bandwidth when servers on opposite switches need to communicate. For example, a Three Hypervisor cluster with 10 Gbps infrastructure NICs requires a 20 Gbps inter-switch link.

warning

When using switch stacking ensure that the configuration also supports In-Service Software Upgrade (ISSU). Rebooting both switches simultaneously will cause a cluster outage.

Disks

The following disk types are supported:

• Enterprise SSD and NVMe drives
• SSD and NVMe drives with DRAM cache
• HCL certified models (Enterprise Flex)

The following disk types may be used in additional disk pools for virtual disks that are written to infrequently, such as file servers. They should not be used for VM boot volumes:

• Nearline SAS HDDs
• Server/NAS SATA HDDs

The following disk types are excluded:

• DRAMless SSDs
• SMR (Shingled Magnetic Recording) HDDs
• Desktop and consumer-grade HDDs

Additional Considerations

When selecting servers for your cluster, consider the following:

• Remote management — Servers with out-of-band management interfaces (iLO, iDRAC, or similar IPMI) make installation and ongoing maintenance significantly easier, especially for headless or remote deployments.
• Network adaptor count — A cluster requires a minimum of 4 network adaptors per server (2 for infrastructure, 2 for VM networking). Ensure your servers have enough NIC ports or expansion slots.
• Disk capacity — Plan for both the boot device and additional data disks for VM storage. More disk slots allow for greater storage flexibility and capacity.

Support Policy

Support Tiers

Gallium classifies hardware into four support tiers:

• Certified — Hardware tested in the Gallium lab in partnership with the vendor. The certified hardware program is available to Enterprise Flex customers.
• Supported — Hardware that Gallium will ensure works through the support channel.
• Compatible — Hardware that is expected to work, but support options may be limited if issues arise.
• Excluded — known incompatible or obsolete, not supported and may be blocked

Support Life

Gallium ends support for a device when the vendor reaches end-of-life (EOL) for that product, unless the device is covered under the Extended Support program. These devices will be classifid as compataible after that date.

Extended Support

Gallium offers extended support for select hardware to enable purchases from the refurbished channel. These models are extensively tested on an ongoing basis in our lab. The following server models are currently covered under extended support for new cluster deployments:

• HPE ProLiant DL360 Gen10 / DL380 Gen10
• Dell PowerEdge R640 / R740
• Cisco UCS C220 M5 / C240 M5
• Cisco UCS B200 M5

Extended support hardware will continue to be supported for at least 3 years from when the cluster license is ordered. We will notify affected customers when an end of support date is announced.

Homelab

For homelab and evaluation use, the recommended hardware is the Minisforum MS-01, MS-A1, and MS-A2 line.

Gallium provides best-effort support for homelab hardware and will not investigate issues on obsolete equipment.