Re: [Qemu-discuss] Increasing speed of guest OS

[邓尧]

IMO, you should refactor your test. If regression test needs as long as 6 hours to complete, the test itself is a problem.

A simple and intuitive approach is to split the test cases into multiple groups, and run the groups in parallel on different servers/VMs

On Thu, Jul 31, 2014 at 6:09 AM, Tony Su <address@hidden> wrote:

 If your main objective is performance, I don't think that there is
any question you should be considering LXC to eliminate overhead.
The reason why LXC is the ultimate in performance is because the
Guests run in a bare metal environment, resources are not virtualized
but simply isolated from other "Guests." and the Host.

QEMU provides features others don't, primarily various emulation but
there is a cost to that.
Paravirtualization (ie implementing KVM) is important but still exacts
a small price compared to LXC.

Tony

On Wed, Jul 30, 2014 at 9:06 AM, Dale R. Worley <address@hidden> wrote:
>> From: chester tinemas <address@hidden>
>>
>> I am trying to increase the speed at which a guest runs in Qemu, so that I
>> can run a regression test on software within the guest OS in a shorter
>> amount of time. Hopefully this would allow a test that would take 6 hours
>> normally take less when run inside qemu. Is this possible using qemu or
>> other software, I've had a look into operators such as -icount but this
>> doesn't seem to be exactly what I'm looking for.
>
> The first step is to ensure that Qemu is using KVM so that the bulk of
> the guest's instructions are executed directly by the hardware.
>
> Of course, Qemu has overheads, so the guest will never run as fast in
> Qemu as it would if it was running directly on the hardware.
>
> If I understand you correctly, you want to make the guest run faster
> than it would if it ran on hardware.  The only way to do that is to
> have Qemu itself running on hardware that is much faster than the
> hardware the guest normally runs on.
>
> Dale
>