[PVE-User] KVM vs Container - Why should one use KVM?

[Daniel Pittman]

Bruce B <bruceb444 at gmail.com> writes:

> I have loaded a few KVM's and a few Container VMs. I don't like how the KVM
> shows using so much of the RAM.

You could make that look less favourable for the OpenVM container by
allocating less than 4GB of RAM to the KVM; we have plenty running in 256MB or
512MB without trouble.  However:

> Comparison below:
>
> KVM = 3.22GB RAM 
> CPU 2%
>
> VM   = 16MB RAM
> CPU = 0%
>
> Both above are running CentOS and Asterisk. So, I am wondering why would
> anyone run in KVM mode?

For Asterisk?  You would have to be happy with crappy performance[1] if you
ran that in a KVM.  It requires very solid timing, and pretty much any
emulated hardware virtual machine will fail on that front.

In the general case:

> Are there any stability?

Some.

> reliability

Some.

> or limitations to Container VM that would restrain one from using a
> Container?

Yes.  There are a whole bunch of things that OpenVZ does not virtualize, so
which can't be used inside the VM, all related to the fact that it needs
specific virtualization support for every kernel feature exposed inside the
container.

OTOH, the same answers apply in the other direction, so YMMV.


Additionally, KVM and OpenVZ containers both have had stability bugs and
certainly have weaknesses in the uncommon or new code they exercise that make
either of them a risky choice for some applications.


> Aside from limitations like one MUST run an .iso to install the OS, are
> there any other features that a container VM can't do that a KVM can do?

Yes.  You can't run a different kernel.  You can't use features of the same
kernel that don't have specific virtualization support.

These are not *common* limitations, but they are real.

Regards,
        Daniel

Footnotes: 
[1]  ...which is as much the fault of Asterisk as the container.

-- 
✣ Daniel Pittman            ✉ daniel at rimspace.net            ☎ +61 401 155 707
               ♽ made with 100 percent post-consumer electrons