This is an archived copy of the Xen.org mailing list, which we have
preserved to ensure that existing links to archives are not broken.
The live archive, which contains the latest emails, can be found at
http://lists.xen.org/
Actually, i would be glad to be wrong about performance numbers. Following bellow is how i set up 2.6.34.7 as bzImage under Xen 4.0.1 on top of Ubuntu 10.10 ( Server or Desktop).
--- On Thu, 10/21/10, Boris Derzhavets <bderzhavets@xxxxxxxxx> wrote:
From: Boris Derzhavets <bderzhavets@xxxxxxxxx> Subject: [Xen-users] Re: [Fedora-xen] another xen build and pvops kernel To: "Pasi Kärkkäinen" <pasik@xxxxxx> Cc: xen@xxxxxxxxxxxxxxxxxxxxxxx, xen-users@xxxxxxxxxxxxxxxxxxx Date: Thursday, October 21, 2010, 1:49 AM
> - How did you measure this performance difference? Database server running on PV DomU with intensive I/O. Workload connected with loading and unloading tables to/from ASCII files residing on the guest's image. > - Did you make sure kernel configuration is similar for both kernel types? > Many pvops .configs have various debugging options enabled that kill the performance..
It's my personal impression. I am not an expert in PVOPS config tuning, but try to avoid debugging options.
Boris.
--- On Wed, 10/20/10, Pasi Kärkkäinen <pasik@xxxxxx> wrote:
From: Pasi Kärkkäinen
<pasik@xxxxxx> Subject: Re: [Fedora-xen] another xen build and pvops kernel To: "Boris Derzhavets" <bderzhavets@xxxxxxxxx> Cc: xen@xxxxxxxxxxxxxxxxxxxxxxx, xen-users@xxxxxxxxxxxxxxxxxxx Date: Wednesday, October 20, 2010, 1:42 PM
On Mon, Oct 18, 2010 at 07:49:29AM -0700, Boris Derzhavets wrote: > Just wondering why [1]kernel-2.6.32.23-170.xendom0.fc12.src.rpm cannot > make itself > into F14. It seems pretty stable. Why it cannot be just a package > available via "yum install" . Anyway Fedora's Libvirt has the best > compatibility with Xen 4.0.1. > It's not too late switch back to Xen with PVOPS kernel. XenLinux 2.6.34.7 > aka Suse > is really dangerous . It outperforms pvops kernel about 15-20 % under Xen >
4.0.1. >
-
How did you measure this performance difference? - Did you make sure kernel configuration is similar for both kernel types? Many pvops .configs have various debugging options enabled that kill the performance..