|
|
|
|
|
|
|
|
|
|
xen-users
Re: [Xen-users] Kernel with Xen and perormance?
Jean-Eric Cuendet wrote:
What is the difference between a kernel with Xen but not used (Dom0
node) and a kernel without Xen code at all?
For distrib, it could be very important that we have a standard kernel
*with* Xen even if 95% of users don't use it (if the perf impact is null
or low) so the 5% of the user that need/want Xen can use it out of the
box without recompiling.
I don't believe that would make sense. None of the distributions that I
use (SuSE, Fedora, Ubuntu currently) come with only one kernel. All
kernels are built for a specific architecture (i386, i586, i686, amd64,
ppc, etc.) though some happen to be backward compatible.
Most of the distributions will detect Athlon vs. Xeon, SMP versus UP,
large vs. small amount of memory, etc. a kernel that is tweaked for that
configuration will be installed (usually without the user having to
specify).
A Xen kernel will not run on any hardware that does not include Xen,
perhaps the Xen kernel code could be made to be backwards compatible
(run-time detection of Xen vs. non-Xen) but it doesn't sound worthwhile
to me.
So, is there a perf penalty if Xen is in the kernel?
To run a Xen kernel you must also run the Xen hypervisor. There is
overhead but it is generally very low especially when running only a
single domain as you are suggesting.
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
|
|
|
|