A hypervisor is software that creates and runs virtual machines. It sits between a physical server’s hardware and the virtual machines on it, giving each one virtual processors, memory, disks and network cards, and deciding how they share the real resources. It also keeps the virtual machines isolated from one another, so a fault or compromise in one should not affect the others. The hypervisor is the core of server virtualization platforms, private clouds and most public cloud compute services.
At a glance
- The hypervisor creates, runs and isolates virtual machines (VMs) on a physical host.
- Type 1 hypervisors run directly on hardware and dominate data centers; Type 2 run on top of a desktop operating system.
- In production it comes with management software for clusters, live migration, high availability and templates.
- Licensing, support and ecosystem compatibility drive most hypervisor buying decisions.
- Because it controls every VM on a host, the hypervisor and its management console need strong protection.
What problem it solves
Operating systems expect to own the hardware they run on. Without a hypervisor, one physical server can safely run one operating system, which leaves most servers underused. The hypervisor solves that by presenting each operating system with its own virtual hardware and arbitrating access to the real thing. That makes virtualization possible: consolidation, fast provisioning, snapshots, live migration and automated restart after hardware failure.
For buyers, the hypervisor is also a platform decision. Backup software, disaster recovery tools, monitoring, automation scripts and staff skills tend to be built around one hypervisor, so changing it is a project, not a setting.
How it works
Placement. A Type 1 hypervisor is installed on the server in place of a general-purpose operating system and boots first. A Type 2 hypervisor runs as an application inside Windows, macOS or Linux. Some designs, such as kernel-based virtualization in Linux, turn a general operating system into a Type 1-style hypervisor.
Resource scheduling. The hypervisor schedules virtual processors onto physical cores, allocates memory, and passes disk and network operations to physical devices or to virtual switches and storage layers. Modern processors include hardware features that let it do this efficiently and enforce isolation.
Cluster management. A management layer pools multiple hosts so administrators can move running VMs between them, restart VMs after a host failure, balance load, and apply updates host by host. Hyper-converged products add distributed storage to the same cluster.
Options. Commercial hypervisors, ones built into server operating systems and open-source hypervisors (often bundled into commercially supported platforms) are all widely used. They differ in licensing model, management tools, support and third-party integrations more than in basic capability.
If you are weighing a hypervisor change against a hosted platform, our hyper-converged infrastructure solution page outlines the trade-offs.
When it matters for buyers
- At a virtualization licensing renewal. Following the acquisition of a leading virtualization vendor, changes to its packaging and licensing led many organizations to evaluate alternatives. Compare the full cost of moving, including migration labor, retraining and backup and monitoring changes, against staying.
- After a merger or acquisition. Two environments often run different hypervisors; standardizing reduces tools and skills to maintain.
- When moving to public cloud or a hosted private cloud. In Infrastructure as a Service (IaaS) the provider runs the hypervisor, so you stop licensing and patching it, but you also lose some control.
- When security reviews flag the management plane. Hypervisor management consoles are a known ransomware target; patching and access isolation should be part of the review.
- When a workload performs poorly or is licensed per physical core. Some teams move it to bare metal instead.
Questions to ask vendors
- How is the hypervisor licensed (per core, per socket, per host or bundled in a platform subscription), and what is the minimum purchase?
- What tools and services do you provide to migrate VMs from our current hypervisor, and what is typically left for us to do?
- Which backup, disaster recovery and monitoring products integrate with your platform today?
- How long is each release supported, and how are security patches delivered?
- What high-availability and live-migration features are included at our licensing level?
- How do we secure and isolate the management interface?
- What is the price protection at renewal, and what notice do you give for licensing changes?
How it differs from bare metal
The terms overlap because a Type 1 hypervisor is often called a “bare-metal hypervisor”, meaning it is installed directly on the hardware. That is different from a bare-metal server or service, where your operating system runs directly on a dedicated machine with no provider hypervisor at all. You can install your own hypervisor on a bare-metal server, which is a common way to build a private cloud on rented hardware. Hypervisors also differ from container platforms, which share one operating system among applications instead of virtualizing hardware.