Skip to content
kryptonservers

What KVM virtualization is, and why it matters on a VPS

KVM gives every VPS its own kernel, memory and disk. What that means in practice, how it differs from container VPS, and what it does not protect you from.

Updated 20 November 2026

“VPS” describes what you buy, a slice of a server with root access, but not how the slice is made. Two VPS with the same RAM and CPU on paper can behave very differently depending on the virtualization underneath. KVM is the one that gives you a whole machine.

KVM in one paragraph

KVM (Kernel-based Virtual Machine) is a hypervisor built into the Linux kernel. With help from the processor’s virtualization extensions (Intel VT-x, AMD-V), it lets one physical server run many virtual machines, each believing it has its own hardware: CPU cores, memory, disks and network cards. Every virtual machine boots its own operating system with its own kernel.

What “your own kernel” gives you

  • Any operating system. Linux of any distribution and version, Windows Server, BSD, or an ISO you bring.
  • Kernel features. WireGuard, custom sysctl tuning, kernel modules, eBPF, a newer or older kernel than the host.
  • Docker and containers inside. Docker, Podman and Kubernetes run as they would on bare metal, with their own networking.
  • Isolation. Other virtual machines on the host cannot read your memory or your disk. A crash or kernel panic in a neighbour’s VM stays in that VM.

The alternative: container VPS

Cheaper VPS are often containers (OpenVZ, LXC). All customers on a host share one Linux kernel; each container is a set of processes with their own view of the file system. That is efficient, which is why it is cheap, but:

  • you cannot change the kernel or load modules, so WireGuard or some VPN setups may not work;
  • Docker may be limited or unavailable;
  • memory and CPU accounting is looser, so a busy neighbour is felt more;
  • a kernel vulnerability on the host is a risk for every customer at once.

If a plan does not say which virtualization it uses, ask. If the answer is “container”, price it as such.

What KVM does not decide

KVM decides isolation. It does not decide how good the server is. Three things matter at least as much:

  • CPU sharing. vCPUs are threads on the host’s processors. Hosts that sell many more vCPUs than they have threads make every VM slower at busy times. Look for honest figures and ask about the host CPU.
  • Storage. Local NVMe is far faster than network storage or SATA SSDs, especially for databases.
  • The network in front. A virtual machine’s network card receives whatever reaches it. Without filtering in front, a small VPS can be taken offline by a flood far smaller than its port speed, because it runs out of CPU first.

KVM and DDoS protection

Because each KVM machine has its own network stack, protection works best before traffic reaches the VM at all. On our VPS, an XDP program in the host’s network driver drops floods and answers new TCP connections with a SYN challenge, so your VM’s CPU is spent on real traffic. More in how DDoS protection works.

When a VPS is not enough

KVM shares a physical server fairly, but it still shares it. If your workload needs every core all the time, a committed network port, or hundreds of gigabytes of RAM, a dedicated server is the honest answer. The trade-offs are in VPS or dedicated server.

Questions about your own setup?

Message @kryptonservers