Re: [Xen-users] HVM Disk performance

["Henning Sprang" <henning_sprang@xxxxxx>]

On 1/23/07, Karl Stevens <karl@xxxxxxxxxxxxx> wrote:
> [...]
> We're trying to install a legacy system (RH7.2) into a VM, but have run
> into a performance issue with disk access.  I understand from the post
> here: http://marc.theaimsgroup.com/?l=xen-users&m=115745154318263&w=2
> that disk access under HVM is quite slow due to the QEmu hardware emulation.
>
> I've read paravirt drivers, and other work being done to improve
> performance in this area, but haven't been able to track down any usable
> code (Rusty Russel has some patches at
> http://ozlabs.org/~rusty/paravirt/ , but it isn't immediately clear what
> it should be applied against.)
AFAIK this is about integrating paravirtualization into the mainstream kernel.
This is more about getting Xen (and other virtualization developer's)
Kernel Patches into the main Kernel source.
I don't think this has to do with using paravirtual drivers in HVM Domains.
Ask Rusty.

Unfortunately, I also don't know/understand yet how using paravirtual
drivers in HVM Domains work, but this technique is even mentioned in
some fine print from XenSource for running windows, because this seems
to be the only way to get the much advertised windows bare metal
performance (and I'm not so impressed by the fact, that their
marketing claims hide the fact that this is only possible with these
paravirtual drivers when claiming the bare metal performance).

I'd be happy to hear where any docs can be found about this, as well
for linux as for windows (where the paravirt drivers might be only
available for xenEnterprise users?!).
Henning

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users