Skip to content

Technology

A look inside our infrastructure.

AMD EPYC · NVMe SSD · ECC RAM · 2× 25 Gbit/s

Hardware

Our server cluster

We exclusively run current enterprise hardware from established vendors. The cluster is built around the AMD Milan generation: AMD EPYC 7443 with up to 24 cores per socket, ECC-Registered Samsung memory at 3200 MHz and Datacenter NVMe SSDs for minimal latencies. Every host is connected redundantly to our network, monitored continuously and protected at every layer. The result: consistent performance, even under load.

CPU
AMD EPYC 7443
RAM
Samsung DDR4 ECC · 3200 MHz
Storage
Samsung Datacenter NVMe
Network
2 × 25 Gbit/s
AMD EPYC Hostsystem — Maincubes FRA01

AMD EPYC 7443 · Maincubes FRA01

Technical Specifications

Host System Specifications

Real cluster values — no marketing numbers.

CPU

AMD EPYC 7443

24 cores / 48 threads, Milan generation

RAM

Samsung ECC-Reg. DDR4

3200 MHz, Registered ECC

NVMe

Samsung Datacenter NVMe

PCIe 4.0, low latency

NIC

2 × 25 Gbit/s

Arista switch, bonded redundant

Virtualisation

KVM / Proxmox VE

Hardware virtualisation, dedicated kernel

Storage

Ceph — 3-way replication

Distributed storage, no single point of failure

Software-defined Storage

CEPH Distributed Storage

Storage does not live on a single machine. It is spread across our entire cluster. Ceph replicates every block of data three times across separate hosts (3-way replication). If a storage node fails, another one takes over automatically, without downtime and without data loss. Horizontal scalability lets us grow capacity without standstill; new hardware is integrated during operation.

More about Ceph
Ceph Distributed Storage

Network

Redundant 25 Gbit/s Infrastructure

Every host has two physical uplinks. If one port fails, the second takes over seamlessly.

Host uplink
2 × 25 Gbit/s
Switch vendor
Arista Networks
DDoS protection
Arbor Networks — always active
Network transit
DE-CIX Frankfurt
Maincubes FRA01 — Serverrack

Software Stack

Proven Enterprise Software

Every component is production-hardened and deployed across the full cluster.

Proxmox

Proxmox VE

KVM virtualisation platform with live migration and HA clustering.

Ceph

Ceph Storage

Software-defined storage — 3-way replication across all storage nodes.

Arista

Arista Networks

High-performance switching with 25 Gbit/s uplinks per host.

AMD EPYC

AMD EPYC

Milan generation — high core density, AVX-512, secure enclaves.

Proxmox VE
AMD EPYC
Arista Networks
Samsung
Arbor Networks

FAQ

Infrastructure Questions

What is the difference between a vServer and a root server?
There is none. Every vServer is a fully KVM-virtualised root server with its own kernel and root access — no container, no shared resources.
Why Ceph instead of local storage?
Local storage means: if one disk fails, the data is gone. Ceph replicates every block three times across separate hosts. Even a complete node failure keeps all VMs running.
What does KVM virtualisation mean technically?
KVM (Kernel-based Virtual Machine) is hardware virtualisation built into the Linux kernel. Your VM gets its own kernel, dedicated CPU access and isolated RAM — no OS-level overselling.
How are the hosts connected to the network?
Every host has two physical 25 Gbit/s uplinks to our Arista switches. Both links are actively bonded — if one fails, the second takes over without interruption.
Where are your servers physically located?
In Maincubes FRA01 and NTT FRA2, both in Frankfurt am Main — a few kilometres from DE-CIX, one of the world's largest internet exchanges.

More questions? Contact us!

If you have any questions, we are happy to help via email.

  • No contract · Prepaid · Online in minutes