Short answer: virtualisation (virtualization in US spelling) is the technique of creating a virtual, software-defined version of something physical — a computer, a disk, a network — so that one physical resource can be split between many isolated users, or many physical resources can be pooled into one. Its best-known form is server virtualisation: a hypervisor divides one physical server into several virtual machines, each running its own operating system as if it had the hardware to itself. Virtual private servers, cloud instances and most corporate data centres run on it.
How does virtualisation work?
A layer of software sits between the physical resource and whatever uses it, and presents a convincing imitation. For servers that layer is the hypervisor: it hands each virtual machine virtual CPUs, memory, disks and network cards, maps them onto the real hardware, and keeps the machines from seeing or affecting each other. The guest operating system runs mostly directly on the processor, and modern CPUs (Intel VT-x, AMD-V) have instructions that make this fast and safe.
What types of virtualisation are there?
| Type | What is virtualised | Examples |
|---|---|---|
| Server (hardware) virtualisation | A whole computer, as virtual machines | KVM, Hyper-V, VMware ESXi, Xen |
| Operating-system-level | An isolated user space on a shared kernel — containers | Docker, LXC, Podman, OpenVZ |
| Desktop virtualisation (VDI) | Desktops running in the data centre, used remotely | Remote Desktop Services, Citrix, VMware Horizon |
| Storage virtualisation | Disks pooled and presented as virtual volumes | LVM, ZFS, Ceph, SAN arrays |
| Network virtualisation | Switches, networks and routing in software | VLANs, VXLAN, Open vSwitch, SDN |
| Application virtualisation | A program packaged with its environment | App-V, Flatpak sandboxes |
| GPU virtualisation | One graphics card shared by several VMs | NVIDIA vGPU, SR-IOV, GPU partitioning |
In hosting, the first two matter most. A VPS is either a full virtual machine or a container sold under the same name, and the difference decides what you can run on it.
Full, para- and hardware-assisted virtualisation
- Full virtualisation runs an unmodified guest operating system, which does not know it is virtual. Before CPUs helped, this needed tricks such as rewriting guest code on the fly (binary translation, VMware’s original approach).
- Paravirtualisation modifies the guest to cooperate: instead of touching hardware, it calls the hypervisor. Xen made this practical in 2003. Today it survives mainly in drivers — virtio on KVM, the VMBus drivers on Hyper-V — which give guests fast disk and network I/O.
- Hardware-assisted virtualisation uses CPU extensions (Intel VT-x and AMD-V from the mid-2000s, plus second-level address translation for memory) so an unmodified guest runs at near-native speed. Every current server hypervisor relies on it.
Virtual machines vs containers
| Virtual machine | Container | |
|---|---|---|
| Kernel | Its own | Shared with the host |
| Operating system | Any: Linux, Windows, BSD | The host's kernel; Linux containers on Linux |
| Isolation | Strong — the hypervisor boundary | Weaker — namespaces and cgroups in one kernel |
| Start-up | Seconds to a minute (it boots) | Milliseconds to seconds (a process starts) |
| Overhead | A full OS per VM | Only the application and its libraries |
| Kernel modules, custom kernels | Yes | No |
| Typical use | VPS, cloud instances, mixed operating systems | Packaging and deploying applications |
They are not rivals. The common pattern is containers inside virtual machines: the VM provides the hard isolation boundary and a kernel you control, and Docker or Kubernetes packages the applications on top. That is also why a container VPS cannot run Docker reliably — the VPS is already a container. The details for VPS buyers are in KVM vs OpenVZ.
Which hypervisors are used?
For servers, the field is small. KVM, part of the Linux kernel, runs most VPS hosting and much of the public cloud, often under platforms such as Proxmox VE or OpenStack. Hyper-V comes with Windows Server and underpins Azure. VMware ESXi dominates traditional enterprise data centres, and Xen lives on in XCP-ng and specialised systems. On desktops, VirtualBox and VMware Workstation run VMs as ordinary applications.
A short history
- 1960s–70s: IBM builds virtual machines for its mainframes — CP-67, then VM/370 in 1972 — so many users can share one expensive computer, each with their own operating system.
- 1999: VMware Workstation brings virtual machines to ordinary x86 PCs, despite x86 not being designed for it.
- 2003: Xen makes paravirtualised servers practical and becomes the basis of early public clouds.
- 2005–2006: Intel VT-x and AMD-V add hardware support for virtualisation to x86 processors.
- 2007: KVM ships in Linux 2.6.20, turning every Linux system into a potential hypervisor.
- 2013: Docker makes Linux containers easy to use, and OS-level virtualisation goes mainstream.
What are the benefits and the costs?
- Consolidation: one well-used server replaces many idle ones — less hardware, power and space.
- Isolation: a crash or compromise in one VM stays in that VM.
- Speed: a new server is minutes away, not a hardware order.
- Snapshots and migration: a VM is a set of files and settings that can be copied, rolled back and moved between hosts, sometimes while running.
- Hardware independence: the guest sees standard virtual devices, so it does not care which physical server it lands on.
The costs are real too:
- Shared resources: VMs on one host compete for CPU time, memory bandwidth and disk I/O; a busy neighbour shows up as CPU steal time (
stintop). - One host, many failures: when a host fails, every VM on it does, unless the platform restarts them elsewhere.
- A thin overhead: small with hardware assistance, but not zero, especially for heavy I/O.
- Side channels: CPU flaws such as Spectre showed that VMs on shared hardware can leak information in rare cases, which hosts mitigate with microcode and kernel updates.
What does virtualisation change for the people running servers?
A great deal of day-to-day operations follows from a VM being software. A provider can move a running VM to another host before hardware maintenance (live migration), take a consistent snapshot before an upgrade, clone a template to deliver a new server in minutes, and restart VMs elsewhere when a host fails. For the customer inside the VM, three things follow: the server’s hardware can change underneath without notice, so nothing should depend on a specific physical machine; performance is shared, so measure it rather than read it off a spec sheet; and the provider’s snapshots are not your backups — they live on the same platform as the VM.
How do you know whether your server is virtual?
systemd-detect-virt # kvm, microsoft, vmware, xen, lxc, openvz, docker — or none
lscpu | grep -i hypervisorOn Windows, systeminfo reports “A hypervisor has been detected” and names the virtual platform as the system model. none from systemd-detect-virt means bare metal.
Virtualisation at VMHeaven
Every VMHeaven VPS is a full KVM virtual machine with its own kernel, never a container: Standard KVM, Hi-CPU KVM, DMCA Ignored KVM and Pentesting KVM, with root access and the freedom to run Docker, WireGuard or your own kernel modules. If you want to run the virtualisation layer yourself — Proxmox VE, plain KVM or Hyper-V — a dedicated server gives you the hardware with no hypervisor of ours on it.
Frequently asked
What is the main purpose of virtualization?
To use hardware better and keep workloads apart: one physical server runs many isolated virtual machines, each with its own operating system, instead of many under-used servers each running one.
Is virtualization the same as cloud computing?
No. Virtualization is the technology that splits hardware into virtual machines; cloud computing is a way of renting resources on demand, which is usually built on virtualization plus automation and billing.
Are containers a type of virtualization?
Yes, operating-system-level virtualization. Containers isolate processes on one shared kernel instead of running separate kernels, which makes them lighter than virtual machines but less isolated.