Re: [Qemu-devel] [RESEND][PATCH 0/3] Fix guest time drift under heavy load.

[Gleb Natapov]

On Wed, Nov 05, 2008 at 02:45:05PM +0200, Dor Laor wrote:
>>> As I've said repeatedly in the past, any sort of time drift fixes 
>>> needs  to have a lot of data posted with it that is repeatable.
>>>
>>> How much does this improve things with Windows?     
>> The time drift is eliminated. If there is a spike in a load time may
>> slow down, but after that it catches up (this happens only during very
>> high loads though).
>>
>>   
> Gleb, can you please provide more details:
> - What's the host's kernel version exactly (including the high-res, dyn  
> tick configured)
Kernel 2.6.25. Qemu uses dynticks (CLOCK_REALTIME) timer. But why
high-res timers should help fix time-drift problem at all? Time drift
happens because guest doesn't get enough time to run, if host timer will
be more precise this will not magically provide guest with more CPU time
to handle them.

> - What's the windows version? Is it standard HAL (pit) or ACPI (rtc) or  
> both?
In my case this what windows2003 with ACPI HAL.

> - The detailed scenario you use (example: I copied the entire c:/windows  
> directory, etc)
Just run one guest in qemu on otherwise un-loaded host and copy entire
c:/windows directory somewhere.

> - Without the patch, what the time drift after x seconds on the host.
After 15 minutes time drift was 5 minutes. But this is with write through
disk cache (default now). With write back cache situation is much
better. After 5 minutes the drift was 8 secs.

> - With the patch, is there a drift? Is there increased cpu consumption, etc
There is no drift with the patch. CPU consumption is 100 with cache
write back so it can be increased any more and with cache write through
qemu doing mostly IO. I am sure that copying will take more time with
the patch since more time interrupts will be handled by guest.

--
                        Gleb.