Tell HN: I hate the existing classification of hypervisors

3 points by tryauuum ↗ HN
Maybe the reason for this is that I'm undereducated, but still.

If you google comparison of hypervisors you will inevitably see "type 1" and "type 2" hypervisors being mentioned. Which is a distinction made in 70s. The idea behind is that with "type 1" the hypervisor is launched on bare metal and with "type 2" first a normal, common OS is launched, and hypervisor is launched inside it.

This is a completely pointless distinction nowadays I think.

If you want a VM with a best performance, it doesn't matter what hypervisor will you use — xen, kvm, esxi. Ultimately, your VM will use hardware virtualization extensions which any modern CPU provides.

The distinction that xen is "type 1" and kvm is "type 2" because they boot differently is completely pointless, who cares how the hypervisor boots if the end result is the same?

Note: I know you don't have to enable cpu hardware virtualization with xen, so the end result can be different. But the point stands: the classification is pointless, why care how the system boots, you should care how it runs VMs.

3 comments

[ 2.7 ms ] story [ 20.9 ms ] thread
I consider it a valid classification, but it looks like you're making assumptions that one type is "better". Your own observations disprove that.

Side note: I swore that KVM was type 1. It's inside the kernel itself so it can reallocate hardware to/from guests and the host. And it's not a separate program like Virtualbox.

this is another reason to dislike this classification, it's not clear enough. KVM can be defined as type 1 or type 2, depending on what you think matters. They do mention in wikipedia page on Hypervisor -- https://en.wikipedia.org/wiki/Hypervisor
(comment deleted)