Re: Xen vs KVM
On 30/03/2012 00:32, francis picabia wrote:
On Wed, Mar 28, 2012 at 1:44 PM, Hilco Wijbenga
On 28 March 2012 06:43, Aaron Toponce<email@example.com> wrote:
On Wed, Mar 28, 2012 at 09:35:25AM +0100, Jon Dowland wrote:
For me, it became yesterday's technology when it became apparent that
the hypervisor model (putting an entirely new kernel between Linux and
the hardware) created all sorts of performance problems, and neglected
the decades of work that had gone into the Linux network stack, amongst
other parts. Increasingly ugly hacks were (are) needed to pass through
to the privileged domain, all of which is totally unnecessary with the
KVM model, where the (much more) tried and tested Linux kernel goes on
the bottom of the pile.
Can you expound on these "ugly hacks"? The Xen kernel is a full type-I
hypervisor, with unfettered access to the hardware. The dom0 presents the
virtualized hardware to the domU guests. Using Xen HVM, the presentation
uses Qemu, which is exactly the same for KVM.
You might both be interested in the PDF linked to at the bottom of
. It explains why Qubes OS went with Xen and not KVM. I thought it
was quite interesting (I used to be firmly in the KVM camp, now I'm
not sure any more. :-) ) Mind you, their focus is mainly security.
Xen requires a patched kernel. It is unstable. It crashed on
me randomly before I got as far as configuring any VM stuff.
The system which experienced this returned to a standard
Debian kernel and never had a problem again.
KVM is native part of kernel. It is stable. I've been running on several
systems for over a year and no crash.
Both share the same qemu devices and drivers land.
You can read what IBM has to say about key benefits and security
features of kvm...
A big clue is Redhat is dropping xen virtualization going forward.
Kvm will get more development support than xen.
I see no reason to even consider xen.
My Xen kernels (3.x) did not crash on me.
Mr. Teo En Ming (Zhang Enming)