Hardware and platforms

AtomOS runs on x86_64 servers with hardware-assisted virtualization. It is installed on a supported Linux distribution, not shipped as its own OS image. Use this page to validate CPU features, firmware, NICs, disks, and optional GPUs before you run atomos-cli.

Release validation required

The matrices below reflect documented support intent. Always validate AtomOS on your exact server model, firmware level, kernel, and AtomOS release before production. Hardware catalog entries (for example the Red Hat Ecosystem Catalog) describe OS compatibility — they do not certify AtomOS behavior.

Sizing: installer minimum vs production

The bootstrap script (get-atomos.sh) enforces a minimum so the CLI can install packages. That bar is not sufficient for useful virtualization.

Tier	CPU	RAM	Boot disk	Intent
Installer minimum	2 cores	2 GB	Enough for OS + packages	Passes bootstrap checks only
Recommended lab / PoC	8 threads (physical cores or SMT threads per your policy)	16 GB	64 GB minimum	Meaningful VM workloads on a single full node
Production	Sized from guest count and specfile CPU policies	Sized from guest RAM and overhead	64 GB minimum; 512 GB+ when VMs share the same disk as the OS	Follow capacity planning in Planning

Separate boot from guest data

Place guest volumes on dedicated disks, RAID sets, ZFS pools, or storage nodes when possible. Sharing one small OS disk with many VMs exhausts IOPS and space quickly.

Compatibility matrix (summary)

Area	Requirement	Notes
Architecture	x86_64 only	ARM has seen testing; production docs focus on x86_64
CPU virtualization	Intel VT-x or AMD-V enabled in firmware	Required for KVM guests
IOMMU	Intel VT-d or AMD-Vi for passthrough	Required for GPU/PCIe passthrough — GPU / PCI passthrough
RAM	See sizing table	Overcommit policy depends on specfiles and workload
Boot firmware	UEFI or legacy BIOS per distro support	Enable virt features in BIOS/UEFI setup
Boot disk	64 GB minimum (recommended table)	OS, AtomOS packages, logs, local vault metadata
Guest storage	Local disks, mdadm, ZFS, GlusterFS, Quobyte, iSCSI, NFS, etc.	Backend choice drives redundancy — Storage architecture
NIC	At least one; two or more recommended	Separate management, guest bridge, and storage traffic when practical
GPU	PCIe with working IOMMU groups	Passthrough constraints apply; not all GPUs suit all topologies

Supported host operating systems

AtomOS is a framework on top of these distros. Package availability depends on role.

Distribution	Version	compute	storage	full	atomosgui	Notes
AlmaLinux	9.5, 9.6, 9.8	Yes	Yes	Yes	Yes	Recommended for compute and full
Rocky Linux	9.5, 9.6	Yes	Yes	Yes	Yes	RHEL-compatible alternative
Ubuntu	24.04 LTS	No	Yes	No	Yes	Storage (and GUI) only
Debian	12, 13	No	Yes	No	Yes	Storage (and GUI) only
AlmaLinux / Rocky	10.x	—	—	—	—	Not supported
AlmaLinux / Rocky	8.x	Deprecated	Deprecated	Deprecated	—	Plan migration to 9.x

Historical Ubuntu releases

Non-LTS or end-of-life Ubuntu versions may appear in older notes elsewhere. For new deployments use 24.04 LTS for storage-only nodes. Do not treat EOL releases as recommended platforms.

Validate on your patch level

Point releases (for example AlmaLinux 9.8 vs 9.5) can differ in kernel and drivers. Re-run preflight checks after distro upgrades — Upgrades.

Public-cloud bare metal and nested virtualization instances can work when the provider exposes KVM-capable CPUs. Deployments exist on several providers; validate the specific instance type and network path before production.

CPU and specfile constraints

Physical sizing must satisfy specfile pledges, not only a raw vCPU count:

Specfile theme	Host impact
Core vs thread allocation	Full-core pledges consume physical cores faster than oversubscribed thread policies
μarchitecture / ISA flags	AVX-512 and other extensions must exist on the host you match
VDI-style overprovisioning	Lighter pledges allow higher oversubscription; monitor steal time and latency

Document peak pledges and reserve headroom for the host OS, storage services, and cluster traffic.

Preflight checks on the host

Run these on a clean or staging install of your chosen distro before atomos-cli --install:

CPU and virtualization

lscpu
grep -E 'vmx|svm' /proc/cpuinfo

Confirm the CPU flags include vmx (Intel) or svm (AMD). Enable VT-x/AMD-V in firmware if flags are missing.

IOMMU (passthrough)

dmesg | grep -e DMAR -e AMD-Vi

For passthrough, firmware must enable VT-d or AMD-Vi, and the kernel must attach IOMMU groups. Inspect groups with atomos-cli --pci-list after install, or lspci and /sys/kernel/iommu_groups/ during staging.

Memory

free -h

Compare available RAM to the recommended table and your guest budget.

Disks and layout

lsblk -o NAME,SIZE,TYPE,FSTYPE,MOUNTPOINT,MODEL

Identify OS disk vs data disks for vaults. Align mount points with your storage design before creating vaults.

Networking

ip -br link
ip -br addr
ip route

Plan which interface becomes the default bridge (br0) for guests and which stays for management SSH and Electros HTTPS.

PCIe devices (GPU/NIC)

lspci -nnk

Record vendor IDs, drivers, and whether devices are bound to vfio or standard drivers before passthrough experiments.

AtomOS-oriented checks (after CLI install)

sudo atomos-cli --healthcheck
sudo atomos-cli --pci-list
sudo atomos-cli --verify-network

Use --verify-install after role deployment — Install.

Firmware and boot

Setting	Recommendation
Virtualization	Enable Intel VT-x / AMD-V
IOMMU	Enable VT-d / AMD-Vi when using passthrough
Secure Boot	Follow distro and guest policy — Security
Power profile	Prefer performance or balanced profiles that do not throttle cores under sustained virt load
NIC firmware	Keep NIC and HBA firmware current for stability under bridge workloads

NIC and network hardware

• At least one NIC; two or more simplify separation of management, br0 guest traffic, and storage networks.
• 10 GbE or faster is common on storage-heavy nodes; 1 GbE can suffice for labs.
• Verify driver support on AlmaLinux or Rocky for compute/full nodes.

See Planning (management and BMC networks) and Networking.

GPU and accelerators

Passthrough requires compatible IOMMU groups, appropriate drivers on the host, and guest configuration. Some workloads prevent live migration when devices are pinned to guests.

Follow GPU / PCI passthrough. Red Hat catalog entries help for OS-level driver support; AtomOS passthrough still needs on-host validation.

Before you install — checklist

1. Record vendor, model, serial, CPU, RAM, firmware version, disk controllers, NICs, and GPUs.
2. Confirm the platform against the Red Hat Ecosystem Catalog when you use RHEL-compatible distros.
3. Install the target distro version from the matrix above (minimal or server variant per your standard).
4. Apply updates; reboot; re-run lscpu, lsblk, and ip checks.
5. Enable virtualization and IOMMU in firmware if passthrough is in scope.
6. Run a PoC: install AtomOS, create test VMs, exercise storage and networking, reboot, and retest workflows you depend on (Electros registration, backup, optional passthrough).

Next steps

• Planning — topologies and network design
• Install with atomos-cli — bootstrap and roles
• Architecture — how hardware maps to services
• Glossary — terms used in specfiles and pledges