System Atlas
the atlas

Types of virtualization

Virtualization isn't just servers — you can fake the kernel, the network, the disk, and the desktop too.

the gist

Server virtualization comes in two flavours: Type 1 hypervisors run on bare metal (ESXi, Hyper-V, KVM) while Type 2 run hosted on an existing OS (VirtualBox). Above them, OS-level virtualization gives you containers, and the same 'abstract the physical resource' idea also virtualizes networks, storage and desktops. It's one principle applied up and down the stack.

ANALOGY // It's like set design in a theatre: you can build a whole fake house (server VM), reuse one stage for many scenes (containers), rewire the lighting without moving cables (network), or pool all the props into one cupboard everyone draws from (storage).
living diagram01 / 07
One idea — abstract a physical resource into many logical ones — applied up the stack.
01
Desktop / app virtualization
VDI · streamed apps · thin clients
02
Storage virtualization
pool disks → logical volumes · SAN · RAID
03
Network virtualization
VLANs · overlays · SDN
04
OS-level virtualization
containers share one kernel
05
Type 2 hypervisor (hosted)
VirtualBox · runs on a host OS
06
Type 1 hypervisor (bare-metal)
ESXi · Hyper-V · KVM
07
Physical hardware
CPU · RAM · disk · NIC
01

Start from physical hardware

At the bottom sits the real machine — CPU, RAM, disks and network cards. Every form of virtualization is a layer that takes one of these physical resources and presents it as many independent logical ones. The whole subject is variations on that single move at different points in the stack.

02

Type 1 hypervisor (bare-metal)

A Type 1 hypervisor installs directly onto the hardware with no operating system beneath it, so it is itself the base software layer. Because it talks to hardware directly, it is fast, efficient and secure, which is why VMware ESXi, Microsoft Hyper-V and Linux KVM run essentially all serious data-center and cloud virtualization. This is what powers the VMs behind AWS, Azure and GCP.

03

Type 2 hypervisor (hosted)

A Type 2 hypervisor runs as an ordinary application on top of an existing host OS, borrowing that OS's drivers to reach the hardware. That extra layer costs some performance but makes it wonderfully convenient — VirtualBox or VMware Workstation lets you spin up a Linux VM on your Windows or Mac laptop in minutes. It is the tool of choice for development, testing and learning.

04

OS-level virtualization (containers)

Instead of virtualizing hardware, OS-level virtualization shares a single kernel and uses namespaces and cgroups to carve it into isolated user-spaces. Each slice — a container — behaves like its own machine but skips the weight of a full guest OS. This is the Docker/LXC model: much lighter and faster than a hypervisor VM, with a softer isolation boundary.

05

Network virtualization

The same abstraction applies to networking: physical switches, routers and cables are turned into software-defined constructs. VLANs segment one physical network into many logical ones, and overlay networks (VXLAN, SDN) let virtual machines and containers talk over virtual links independent of the underlying wiring. This is what lets a cloud give each tenant its own private network on shared hardware.

06

Storage virtualization

Storage virtualization pools many physical disks — spread across many devices or arrays — into logical volumes that are managed as a single resource. Administrators carve out, resize and migrate storage without users ever knowing which physical drive holds their data. SANs, RAID and cloud block storage all rest on this abstraction.

07

Desktop virtualization (VDI)

At the top, desktop virtualization runs users' entire desktop environments on central servers and streams the screen to lightweight thin clients or any device. IT manages, patches and secures desktops in one place, and a user can pick up the same session from any machine. Application virtualization is a narrower cousin that streams individual apps rather than whole desktops.

the full picture
01

One idea, many places to apply it

Virtualization is the practice of creating a software abstraction of a physical resource so that one real thing can be presented as many independent logical ones — or, sometimes, many things as one. People usually mean "servers" when they say virtualization, but the same principle applies to the CPU and kernel, the network, storage, and even the desktop. Learning the types is really learning where in the stack you can insert this abstraction and what you gain when you do.

The payoff is always the same family of benefits: better utilisation of expensive hardware, isolation between workloads, flexibility to resize and move things without touching cables or drives, and centralised management. What changes from type to type is which physical resource is being faked and how thick the abstraction layer is.

02

Server virtualization: Type 1 vs Type 2 hypervisors

The most familiar form is server (hardware) virtualization, run by a hypervisor that lets multiple virtual machines share one physical host. Hypervisors come in two types, and the difference is simply what sits beneath them.

A Type 1 (bare-metal) hypervisor installs directly onto the hardware — there is no host operating system at all; the hypervisor *is* the base layer. Because it talks to hardware directly, it delivers the best performance, efficiency and security, which is why it dominates data centers and the cloud. VMware ESXi, Microsoft Hyper-V and Linux's KVM are the canonical examples, and they are what actually run the instances you rent from AWS, Azure and GCP.

A Type 2 (hosted) hypervisor runs as an ordinary application on top of an existing OS, using that host's drivers to reach the hardware. The extra indirection costs some performance, but it is enormously convenient for individuals: VirtualBox and VMware Workstation let you run a Linux VM inside Windows, or test three operating systems on one laptop, with a couple of clicks. Rule of thumb — Type 1 for production and servers, Type 2 for desktops, development and experimentation.

03

OS-level virtualization: the container model

Hypervisors virtualize hardware, giving each guest its own kernel. OS-level virtualization takes a lighter approach: it shares the host's single kernel and uses kernel features — Linux namespaces for isolation and cgroups for resource limits — to split that one OS into many isolated user-space environments. Each of those environments is a container.

Because there is no second kernel to boot and no duplicated OS to store, containers start in milliseconds and their images are megabytes rather than gigabytes, so you can pack far more of them onto a host. The trade-off is a softer isolation boundary, since everything shares the same kernel. This is the model behind Docker and LXC, and it sits one rung above the hypervisor in the virtualization stack — often literally, with containers running inside VMs.

04

Beyond compute: network, storage and desktop

The abstraction reaches well past compute. Each of the following virtualizes a different physical resource:

  • Network virtualization turns physical switches, routers and cables into software-defined constructs. VLANs segment one physical network into many logical ones, while overlay networks and software-defined networking (SDN) let VMs and containers communicate over virtual links that are independent of the underlying wiring. This is how a cloud gives every tenant its own private, isolated network on shared physical infrastructure.
  • Storage virtualization pools many physical disks — potentially across many arrays — into logical volumes managed as a single resource. Administrators provision, resize and migrate storage without users knowing or caring which drive holds their bytes. SANs, RAID and cloud block storage all build on this.
  • Desktop virtualization (VDI) runs entire user desktops on central servers and streams them to thin clients or any device, so IT patches and secures desktops in one place and users roam between machines seamlessly. Application virtualization is a narrower variant that streams individual apps rather than a whole desktop.
05

Why the taxonomy matters

Seeing these as one family rather than unrelated products is what makes cloud architecture legible. A modern cloud region is every type stacked together: bare-metal Type 1 hypervisors carve physical servers into VMs; OS-level virtualization packs containers inside those VMs; network virtualization draws each customer a private virtual network across the shared fabric; storage virtualization presents elastic block and object stores backed by vast disk pools; and desktop virtualization delivers managed workspaces on top.

Knowing which type you are dealing with tells you what its boundary guarantees, where its overhead comes from, and what it can and cannot do. A Type 1 hypervisor gives a hard isolation wall but heavy VMs; a container gives speed and density but shares a kernel; network virtualization isolates traffic but not compute. When you can name the layer, you can reason about performance, security and cost for the whole system instead of guessing — which is exactly the skill these interview questions are probing.

what to remember
01

Type 1 (bare-metal) hypervisors run directly on hardware with no host OS — ESXi, Hyper-V, KVM — for high performance in data centers.

02

Type 2 (hosted) hypervisors run as an app on top of a normal OS — VirtualBox, VMware Workstation — great for desktops and testing.

03

OS-level virtualization shares one kernel to create isolated containers instead of full VMs.

04

Network virtualization abstracts physical networking into VLANs, overlays and software-defined networks.

05

Storage virtualization pools many physical disks into logical volumes managed as one.

06

Desktop virtualization (VDI) runs user desktops on central servers and streams them to thin clients.

check yourself

Quick quiz

Q1

What distinguishes a Type 1 from a Type 2 hypervisor?

Q2

Which type of virtualization produces containers?

follow the thread