Free — no signup required

Introduction: The Hypervisor

2 min read

The Software That Ate the World

Before the cloud, if you wanted a server, you bought a physical box. It sat in a rack, ran one operating system, and used maybe 10–15% of its CPU capacity on average. The rest was wasted. Today, nearly every server you interact with is a "Virtual Machine" (VM) — a software-defined computer running inside a physical one.

The magic software that makes this possible is the Hypervisor (also called the Virtual Machine Monitor, or VMM). Think of it like an apartment building manager: the physical hardware is the building, and the hypervisor divides it into individual apartments (VMs), each with its own walls, locks, and utilities. Tenants (guest operating systems) can't see each other's space, and the manager controls who gets what resources.

There are two main types:

  1. Type 1 (Bare Metal): The Hypervisor installs directly on the hardware. There is no "Host OS" in between — the hypervisor is the operating environment.
  2. Examples: VMware ESXi, Microsoft Hyper-V, Xen.
  3. Use Case: Enterprise datacenters, AWS, Google Cloud.
  4. Note on KVM: Linux's KVM (Kernel-based Virtual Machine) is a kernel module that turns the Linux kernel itself into a Type 1 hypervisor. The Linux kernel handles scheduling and memory, while KVM adds hardware virtualization extensions (Intel VT-x / AMD-V). This blurs the Type 1/Type 2 line — KVM is technically Type 1 because it runs in kernel space, but it requires a host Linux OS to boot.

  5. Type 2 (Hosted): The Hypervisor runs as an application inside a normal operating system. The host OS mediates all hardware access.

  6. Examples: VirtualBox, VMware Fusion, Parallels.
  7. Use Case: Running Windows on your Mac, or testing a Linux distribution on your Windows laptop.
  8. Trade-off: Convenient but slower — every hardware call passes through the host OS before reaching the hypervisor.

Key Point: The reason cloud providers use Type 1 hypervisors is raw performance and security isolation. AWS's Nitro hypervisor, for example, offloads networking and storage virtualization to dedicated hardware ASICs, reducing the hypervisor's attack surface to near zero and giving VMs near-native I/O performance. Understanding this architecture explains why "bare metal" cloud instances (no hypervisor at all) exist as a product — some workloads, like high-frequency trading or certain database engines, cannot tolerate even the thin overhead of a modern Type 1 hypervisor.

This is one of 18 chapters

Every chapter is free — Kubernetes, Terraform, SRE, distributed systems, and more — with fast daily review built in.

Sign up free