Dedicated servers

A dedicated server is a whole physical machine: every core, all the memory, the disks and the network port are yours. Servers marked in stock are pre-racked and imaged automatically, which is why they are online within a minute of payment rather than days. This guide covers the families, delivery, out-of-band access, disks, IP addresses, bandwidth and what happens when hardware fails.

Families

All servers use DDR5 ECC memory and NVMe Gen4 system disks. Every location has dedicated servers; the price is the same in all six. Live stock is shown on the dedicated page.

Delivery

Servers marked in stock are already racked, cabled and connected to the provisioning network. When the order is paid from your balance, the machine is wiped, the image you chose is written, the network is configured and the root password is generated — usually within 60 seconds. Custom builds (other CPUs, more memory, specific disk layouts) are quoted through a ticket and take up to 3 business days.

The initial installation uses the whole first NVMe disk with a single root partition. If you want software RAID or a separate data volume prepared for you, open a ticket with the layout before or right after ordering; otherwise set it up yourself from the rescue system as described below.

IPMI and rescue system

Each server has an IPMI/BMC reachable through the dashboard. From there you can power-cycle the machine, open a HTML5 KVM console, read sensor values and mount an ISO as virtual media. IPMI sessions are opened on demand and expire after 4 hours; the BMC itself is on a management network that is not reachable from the internet.

The rescue system boots the server over the network into a Debian-based environment with all disks attached. It is the fastest way to repair a broken bootloader or copy data off a machine that will not boot:

lsblk
mdadm --assemble --scan          # if you use software RAID
mount /dev/nvme0n1p2 /mnt
chroot /mnt /bin/bash

Disks and RAID

Ryzen and EPYC servers ship with two or four NVMe disks; the storage server has twelve 20 TB hard disks plus two 1.92 TB NVMe. We do not install hardware RAID controllers: NVMe performs best directly attached, and software RAID (mdadm, ZFS, btrfs) is transparent and portable across chassis swaps.

To mirror two NVMe disks after a default installation, boot the rescue system, create a degraded RAID 1 on the second disk, copy the system over, boot from it and finish by adding the first disk to the array. For the storage server, ZFS with two RAID-Z2 vdevs of six disks and the NVMe pair as a special device is a proven layout:

zpool create -o ashift=12 tank raidz2 /dev/sd[a-f] raidz2 /dev/sd[g-l]
zpool add tank special mirror /dev/nvme0n1 /dev/nvme1n1

IP addresses

Every server comes with one IPv4 address and a routed /64 IPv6 block. Additional IPv4 addresses can be requested from the dashboard, up to a /29 (8 addresses) per server, with a short justification as required by the RIRs. They are routed to your primary address and can be moved between your servers in the same location. Larger IPv6 blocks (a /56) are available on request, and you can announce your own prefixes over BGP; see Network.

Reverse DNS for every address is editable from the dashboard and propagates within minutes.

Bandwidth and 10 Gbps

Ryzen and storage servers use an unmetered 1 Gbps port: no traffic accounting, no throttling. EPYC and Xeon servers come with a 10 Gbps port and 100 TB (DS-9354P, DS-6526Y) or 200 TB (DS-9554P) of monthly outbound traffic; beyond that the port is limited to 1 Gbps until the next cycle, never billed. A 10 Gbps port can be added to a Ryzen server on request. Ports are never oversubscribed at the rack level, and each site has two 100 Gbps uplinks.

Operating systems

Choose any image from the catalogue — Ubuntu, Debian, Rocky, AlmaLinux, Fedora, Arch, FreeBSD or a one-click app such as Proxmox VE or TrueNAS — or install your own system from an ISO mounted through the IPMI virtual media. Windows Server with the licence included is available on request. Reinstalling is a one-click operation in the dashboard and keeps the IP addresses.

Hardware replacement

We keep a spare pool of disks, memory modules, power supplies and complete chassis in every location. When a component fails, open a ticket (or let our monitoring open it for you — SMART and ECC alerts are watched) and a technician replaces it within 4 hours. If the chassis has to be swapped, your disks are moved to the new machine whenever they are healthy, so the system boots unchanged. Disks are never reused for another customer: a replaced disk is destroyed.

Something missing or wrong on this page? Open a ticket and tell us.