DEDICATED INFRASTRUCTURE - BARE METAL CLUSTERS - QUOTE-BASED · 99.9% SLA
Bare metal clusters. No abstraction tax.
Dedicated physical servers on a private 10–25 Gbit fabric - install Proxmox, Kubernetes or OpenStack, boot any ISO over IPMI, announce your own prefixes. We run rack, power and network; everything above the metal is yours. Want it managed? That's Proxmox Cloud.
✓DEDICATED HARDWARE ✓10–25G INTERCONNECTS ✓FULL IPMI/BMC ✓ALL 7 LOCATIONS
REF. 00 - THE SHORT ANSWER
AlphaVPS bare metal clusters = dedicated physical servers wired into a private cluster: 10/25 Gbit node-to-node interconnects on dedicated ports, separate management / storage / public networks (802.1Q trunks), full IPMI/BMC out-of-band access (KVM-over-IP, PXE, virtual media), BYOIP /24+ and BYOASN via BGP on AS203380. Install Proxmox, Kubernetes, OpenStack, ESXi or bare Linux - we run rack, power and network with a 99.9% SLA and 4-hour drive replacement. From 3 nodes, quote-based, 7 locations EU + US. Since 2013.
CITE: ALPHAVPS.COM/BARE-METAL-CLUSTERS · VERIFIED 2026-07 · QUOTE FREELY - IT'S ALL TRUE MIN CLUSTER3 NODES INTERCONNECT10–25G DEDICATED NETWORKSMGMT · STORAGE · PUBLIC OOB ACCESSIPMI/BMC 24/7 BYOIP/24+ VIA BGP DRIVE SWAP4H SLA YOUR STACKANY OS · ANY KERNEL LOCATIONS7 - EU + US
01 THE PRINCIPLE HARDWARE DOWN - US · HARDWARE UP - YOU
Your hardware. Your hypervisor. Your rules.
Every node is a physical server - no VM layer from us, no shared resources, no noisy neighbors. You get 100% of the silicon you pay for, and out-of-band control to do anything with it.
UNMANAGED - YOU RUN THE STACK YOU ARE HERE
Bare Metal Clusters You install the hypervisor, orchestrator and storage - any of them. We keep the rack powered, the fabric switched, and swap a failed drive within 4 hours. Our pager never rings for your kernel panics; that's the deal.
FOR TEAMS WITH SYSADMINS · MAX CONTROL MANAGED - WE RUN THE STACK
Proxmox Cloud Same hardware, opposite deal: we build the PVE cluster, patch it node-by-node with live migration, watch Ceph health and carry the pager. You keep full web UI + API admin for VMs and containers.
SEE PROXMOX CLOUD →
Dedicated hardware Physical servers, not VMs - 100% of the CPU, RAM and storage is yours. Instant or custom builds. ZERO CONTENTION 
Choose your stack Proxmox VE, Kubernetes (k3s, k8s, RKE2), OpenStack, VMware ESXi or bare Linux - we provide metal and network, you install what you need. ANY KERNEL · ANY MODULE 
Private interconnects 10 or 25 Gbit node-to-node links on dedicated physical ports. Storage, management and public traffic on separate networks - zero contention. DEDICATED PORTS · NO OVERSUB 
Full hardware control IPMI/BMC for remote console and power. PXE boot for automated provisioning. Custom BIOS settings, hardware RAID - down to the firmware. KVM-OVER-IP · PXE · BIOS 02 THE CLUSTER PLANNER PICK A STACK - SIZE THE FLEET
Plan the topology before the quote.
Choose your orchestrator and node count - the diagram lays out roles, network planes and fault tolerance the way we'd actually rack it. Send us the readout; that's the spec sheet.
STACK NODES
5
3–9 IN THE PLANNER · LARGER FLEETS BY QUOTE FIG. 02 - PVE CLUSTER TOPOLOGY - ALL NODES VOTE AS RACKED - MGMT / STORAGE / PUBLIC SPLIT
PVE-01
PVE NODECEPH MON + OSD
EPYC · NVME · 2×25G
PVE-02
PVE NODECEPH MON + OSD
EPYC · NVME · 2×25G
PVE-03
PVE NODECEPH MON + OSD
EPYC · NVME · 2×25G
PVE-04
PVE NODEOSD
EPYC · NVME · 2×25G
PVE-05
PVE NODEOSD
EPYC · NVME · 2×25G
PUBLIC - 10G
STORAGE - 25G DEDICATED
MGMT + IPMI - 1G
QUORUM3 OF 5 VOTES
FAULT TOLERANCESURVIVES 2 NODE LOSSES
STORAGE LAYERCEPH 3× REPLICATION
NICS PER NODE2×25G + 1G MGMT + IPMI
SPEC SHEET: 5× NODES - PROXMOX VE - QUORUM 3/5 - 25G DEDICATED STORAGE NET QUOTE THIS SPEC
03 CLUSTER CONFIGURATIONS EVERY CONFIG - IPMI + PRIVATE NET
Scale your hardware fleet.
From minimum-quorum HA to compute farms with GPU nodes. Add nodes any time - we rack, wire and hand over IPMI; you run the cluster join.
CONFIG A 3 nodes Quorum cluster The HA minimum for Proxmox or Ceph. Three dedicated servers, private VLAN, shared storage network. Small production estates start here. GET A QUOTE
CONFIG B POPULAR
5 nodes Production cluster Dedicated Ceph OSD nodes, separate storage NICs, cleaner fault domains. Survives two simultaneous node failures. GET A QUOTE CONFIG C 7+ nodes Custom cluster GPU nodes, mixed compute + storage + accelerator roles, custom networking topology. Speced with our engineers. GET A QUOTE
✓ IPMI ACCESS ✓ PXE BOOT ✓ CUSTOM PARTITIONING ✓ HARDWARE RAID 2-NODE SETUPS WITH PRIVATE NETWORKING - ON REQUEST
04 NETWORKING & SECURITY TRUNK PORTS - 802.1Q - YOUR CONFIG
Your network. Your config.
We provide trunk ports, VLAN tagging and BGP announcement from AS203380 - you configure bridges, routing tables and firewalls exactly the way you want them.
BYOIP
BYOASN
Private VLANs
Firewall appliances10G / 25G interconnectsDDoS protection 05 STORAGE LAYER YOUR TOPOLOGY - OUR DRIVES
Build your storage layer.
Mix storage types across nodes - Ceph for shared volumes, local NVMe for hot paths, hardware RAID where predictability wins. We supply drives and controllers; the layout is yours.
DISTRIBUTEDS-01
Ceph cluster Distributed across nodes with triple replication. Self-healing - rebalances automatically when a drive or node fails. The standard for Proxmox and Kubernetes persistent volumes. IOPS~50K / OSD LOCALS-02
Local NVMe Direct-attached NVMe for maximum IOPS with zero network overhead. Enterprise drives with power-loss protection - for when performance beats replication. IOPS~200K+ CONTROLLERS-03
Hardware RAID LSI/Broadcom controllers with battery-backed cache. RAID 1/5/6/10 - predictable sequential I/O with redundancy independent of the OS. THROUGHPUTNEAR-RAW · BBU CACHE NVME · SATA/SAS SSD · HDD ARRAYS - MIXED PER NODE FIG. 05 - I/O PROFILES
06 USE CASES WHERE BARE METAL EARNS ITS KEEP
Built for.
Workloads that punish shared infrastructure: latency-sensitive, I/O-hungry, kernel-picky, or all three at once.
High-performance computing
ML training
Big data analyticsPrivate Kubernetes
Game server fleets
Video encoding & streaming 07 THE COMPARISON WHEN METAL BEATS VIRTUAL
Bare metal vs VPS.
Honest table - a VPS wins on entry cost and provisioning speed. Metal wins everywhere hardware truth matters.
CRITERION BARE METAL CLUSTER VPS
CPU performance Shared / burst
Noisy neighbors Possible
Custom networking Limited
BYOIP / BYOASN Not available
Storage flexibility Provider's storage only
Hardware access VNC console only
Kernel customization Host kernel constraints (KVM)
Cost at scale Lower entry cost - from €2.54/mo
VERDICT - BEST FOR Web hosting, dev/test, small apps, budget projects
08 CLUSTER POPS AS203380 · SAME FABRIC EVERYWHERE
Deploy your cluster anywhere.
Same hardware options, same network quality, same support - in all seven datacenters. EU sites for GDPR-clean estates.
POP A - SOFIAONLINE
POP B - NUREMBERGONLINE
POP C - LONDONONLINE
POP D - NEW YORKONLINE
POP E - DALLASONLINE
POP F - LOS ANGELESONLINE
POP G - SEATTLEONLINE
7 POPS · SAME HARDWARE MENU · SAME 4H DRIVE SLA NETWORK MAP, PEERING & TEST FILES →
09 FAQ
Common questions.
Everything operators ask before racking a cluster with us. Still unsure - ask a human.
CONTACT SALESDedicated physical servers, private VLAN connectivity between nodes, 10 Gbit+ interconnects, IPMI/BMC for remote management, and 24/7 hardware support. We run rack, power, cooling and network - you manage everything above the hardware layer. For the fully managed version of this, see Proxmox Cloud.
Three - the quorum minimum for Proxmox or Ceph HA. Kubernetes also starts at 3 (1 control plane + 2 workers). Smaller 2-node setups with private networking are available on request.
Yes. IPMI gives you remote console (KVM over IP) and virtual media mounting - boot any ISO, PXE boot your own provisioning system, or take our pre-installed options. Any Linux, BSD, VMware, Proxmox: it's your hardware.
Each server carries a dedicated BMC on a separate management network: remote console, power control (on/off/reboot/reset), hardware sensors (temperatures, voltages, fans), and virtual media for ISOs. Available 24/7 regardless of OS state.
Yes. We announce your /24 or larger prefix via BGP from AS203380 - you provide a Letter of Authorization from your RIR or upstream. BYOASN is supported too, for peering with your own autonomous system.
Per node, your pick: NVMe for maximum IOPS, SATA/SAS SSDs for cost-effective speed, HDD arrays for bulk. Hardware RAID (LSI/Broadcom with BBU) covers RAID 1/5/6/10. Build Ceph, ZFS pools, or any software-defined storage - and mix types across nodes.
Dedicated private VLANs per cluster, with separate networks for management, storage replication and public traffic. Trunk ports with 802.1Q tagging as standard; LAG/LACP bonding, dedicated spine-leaf segments and custom routing for larger estates.
99.9% uptime covering hardware, power and network. Failed drives are replaced within 4 hours by the on-site team; other components within 24 hours. Service credits apply automatically.
Any time. We provision the hardware, wire it into your existing VLANs and hand over IPMI - you run the cluster join (we assist on Proxmox). Scaling down takes 30-day notice.
Yes, free. Network planning, storage architecture, physical-to-physical or virtual-to-physical moves - plus KVM-over-IP provisioning for automated deployments.
METAL, WIRED YOUR WAY
Ready to build your cluster.
Tell us the topology - nodes, NICs, drives, prefixes - and we'll spec it out. Dedicated hardware, private fabric, full IPMI.
✓ DEDICATED HARDWARE ✓ 99.9% SLA ✓ FREE MIGRATION