What Is Virtualization? Hypervisors and VMs Explained

Every time you launch a cloud server, spin up a test environment or run Windows on a Mac, you are using virtualization. It is one of the most important technologies in modern IT, and also one of the least visible. Most people only notice it when a license bill arrives or a server runs slowly.

The idea is simple: instead of dedicating one physical computer to one job, you split that computer into several independent virtual computers. Each one behaves as if it has its own hardware. In reality, they all share the same physical machine, managed by a thin layer of software called a hypervisor.

This guide explains how that works, the main types of hypervisors and virtualization, and why the whole cloud industry is built on it.

What virtualization means

Virtualization uses software to create an abstraction layer over physical hardware. That layer lets one physical machine be divided into multiple isolated virtual machines, or VMs. Each VM runs its own operating system and applications and acts like a separate computer.

A few terms come up constantly:

• Host: the physical server that provides the hardware.
• Guest: a virtual machine running on the host.
• Hypervisor: the software that creates, runs and isolates the VMs.
• Virtual machine image: a file containing a VM’s disk and configuration, which can be copied, backed up or moved.

The concept is older than most people expect. IBM introduced CP-40, a time-sharing research project for its System/360 mainframes, in 1964. It evolved into CP-67, which let a single real machine appear as many virtual machines. Virtualization moved onto ordinary x86 servers decades later and became mainstream in data centers in the 2000s.

How one server runs many machines

A modern server might have dozens of processor cores and hundreds of gigabytes of memory. Very few applications need all of that at once. Without virtualization, most of that capacity sits idle.

The hypervisor solves this by acting as a traffic controller between the VMs and the hardware:

1. CPU: the hypervisor schedules each VM’s virtual processors onto the physical cores, so many VMs take turns on shared hardware.
2. Memory: each VM gets its own slice of RAM and cannot read another VM’s memory.
3. Storage: each VM sees a virtual disk, usually a file or volume on shared storage.
4. Networking: VMs get virtual network adapters connected to virtual switches, which link to the physical network.

Modern processors include hardware features that help hypervisors do this efficiently and securely, so the performance cost of virtualization for most workloads is small.

Key takeaway: A hypervisor turns one large physical server into many smaller, isolated virtual servers. That is how cloud providers rent you a “server” in seconds without installing new hardware.

Type 1 vs Type 2 hypervisors

Hypervisors come in 2 main types.

KVM is a special case. It is built into the Linux kernel, where it was merged in about 2007, and turns Linux itself into a hypervisor in which each VM runs as a regular Linux process. It is widely used in clouds and in platforms such as Proxmox VE and OpenStack.

Types of virtualization

Server virtualization is the best known, but the same idea applies across IT.

Why virtualization underpins the cloud

When you launch a virtual machine on AWS, Microsoft Azure or Google Cloud, a hypervisor on one of the provider’s physical servers creates it for you. That is what makes infrastructure as a service possible. Our guide to IaaS vs PaaS vs SaaS shows how higher-level services build on top.

AWS publishes detailed documentation on its own design, the Nitro System. It moves networking, storage and management functions onto dedicated hardware cards, leaving a deliberately minimal Nitro Hypervisor whose main jobs are partitioning CPU and memory and assigning virtual devices to each VM. AWS notes that its hypervisor has no networking stack, no general-purpose file system and no interactive shell, which reduces the attack surface.

The hypervisor is also a key dividing line in cloud security. The provider secures the physical hosts and the hypervisor; with virtual machines, you secure the guest operating system and everything inside it. Our explainer on the cloud shared responsibility model covers that split.

Benefits and limits of virtualization

Benefits: - Higher utilization: many workloads share one server instead of each needing its own. - Faster provisioning: a new VM takes minutes, not weeks of hardware procurement. - Easier recovery: VMs can be snapshotted, backed up, replicated and restarted on other hosts. Our guide to backing up business data covers strategy. - Isolation: a crash or compromise in one VM is contained. - Lower costs: fewer physical servers mean less hardware, power, cooling and space. - Legacy support: older operating systems and applications can keep running on new hardware.

Limits: - Overhead: each VM runs a full operating system, which uses memory and storage. - Noisy neighbors: VMs on the same host compete for resources if the host is overloaded. - Single point of failure: if a host fails, all its VMs go down unless you have high availability configured. - Licensing complexity: hypervisor, operating system and application licenses may be priced per core, per host or per VM. - Management sprawl: VMs are so easy to create that unused ones pile up.

How virtual machines compare with containers

Containers are a lighter-weight alternative for many applications. Instead of virtualizing hardware and running a full operating system per VM, containers share the host’s operating system kernel and package only the application and its dependencies. They start faster and use fewer resources, but offer less isolation than VMs.

In practice, the 2 work together. Most containers in the cloud run inside virtual machines. Our comparisons of containers vs virtual machines and Docker vs Kubernetes go deeper, and our explainer on serverless computing covers the next layer of abstraction.

The virtualization market is shifting

If you run your own virtualized data center, the vendor landscape has changed. Broadcom completed its acquisition of VMware on November 22, 2023, and in December 2023 it ended sales of new perpetual VMware licenses in favor of subscriptions. Many organizations have since reviewed their virtualization costs and options, including alternatives such as Microsoft Hyper-V, KVM-based platforms and moving workloads to the public cloud. Some are going the other way and moving steady workloads back to owned hardware, which we cover in our guide to cloud repatriation.

Where to start

If you are new to virtualization or reviewing your setup, use these 5 steps.

1. Experiment on a laptop. Install a free Type 2 hypervisor and create a Linux VM. Seeing a VM boot inside your computer makes the concepts concrete.
2. Inventory what you run. List every VM, its owner, purpose, resources and operating system. Remove anything unused.
3. Check host capacity. Make sure hosts are not overcommitted on memory or CPU, which causes slowdowns across every VM.
4. Review licensing. Confirm how your hypervisor, operating systems and applications are licensed, and whether recent changes affect your costs.
5. Plan for failure. Set up backups, snapshots and, for critical systems, high availability across multiple hosts.

Conclusion

Virtualization is the technique of running many isolated virtual machines on one physical computer, managed by a hypervisor. Type 1 hypervisors run on bare metal and power data centers and clouds; Type 2 hypervisors run on desktops for testing and development. It raises utilization, speeds up provisioning and makes recovery easier, and it is the foundation every major cloud platform is built on. Understanding it makes cloud pricing, security and architecture decisions much easier to reason about.

Sources

• IBM: What is virtualization? — definition, hypervisor types, benefits and CP-40 history
• AWS Whitepaper: The components of the Nitro System — Nitro Hypervisor design and responsibilities
• Red Hat: What is KVM? — KVM as a Linux kernel hypervisor and its history
• Broadcom: Broadcom completes acquisition of VMware — November 22, 2023 completion of the acquisition