OpenVZ Forum: Users » CPUUNITS (yeah...yet another question...*SIGH*)

Home » Mailing lists » Users » CPUUNITS (yeah...yet another question...*SIGH*)

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

CPUUNITS (yeah...yet another question...*SIGH*) [message #14571]

Mon, 02 July 2007 14:56

Michael Portz
Messages: 2
Registered: July 2007

Junior Member

Hi!

Everywhere it is stressed, that CPUUNITS is only relevant for the
relative amount of time a VE has control of the resources. I am
looking for a quite different answer: Does it have an absolute
meaning as well?

E.g. if VE0's CPUUNITS=1000, VE1's CPUUNITS=1000 and VE2's CPUUNITS=1000
(and these are *all* VEs), then they all get the same share of processing
time. The same holds for CPUUNITS=100 for all VEs. But is there any
difference for the values 100 and 1000 respectively?

Does e.g CPUUNITS=100 mean, a VE is swapped out after 100 * c timeunits
for a constant c? Specifically: Does decreasing the values of CPUUNITS
decrease the reaction time of VEs towards interrupts etc.?

Thanks
Michael

Report message to a moderator

Re: CPUUNITS (yeah...yet another question...*SIGH*) [message #14575 is a reply to message #14571]

Mon, 02 July 2007 15:53

dev
Messages: 1693
Registered: September 2005
Location: Moscow

Senior Member

No, 100 is absolutely the same as 1000 in this regard.
CPUUNITs control only how VEs fight for the CPU time and
doesn't affect latency of the reaction which is controled
by HZ and some of sysctls in /proc/sys/kernel and is *bounded*.

Thanks,
Kirill

Michael Portz wrote:
> Hi!
>
> Everywhere it is stressed, that CPUUNITS is only relevant for the
> relative amount of time a VE has control of the resources. I am
> looking for a quite different answer: Does it have an absolute
> meaning as well?
>
> E.g. if VE0's CPUUNITS=1000, VE1's CPUUNITS=1000 and VE2's CPUUNITS=1000
> (and these are *all* VEs), then they all get the same share of processing
> time. The same holds for CPUUNITS=100 for all VEs. But is there any
> difference for the values 100 and 1000 respectively?
>
> Does e.g CPUUNITS=100 mean, a VE is swapped out after 100 * c timeunits
> for a constant c? Specifically: Does decreasing the values of CPUUNITS
> decrease the reaction time of VEs towards interrupts etc.?
>
> Thanks
> Michael
>
>

Report message to a moderator

Re: CPUUNITS (yeah...yet another question...*SIGH*) [message #14581 is a reply to message #14575]

Mon, 02 July 2007 18:28

Michael Portz
Messages: 2
Registered: July 2007

Junior Member

Ah, ok!

Thanks for the quick answer :)
Michael

Kirill Korotaev schrieb:
> No, 100 is absolutely the same as 1000 in this regard.
> CPUUNITs control only how VEs fight for the CPU time and
> doesn't affect latency of the reaction which is controled
> by HZ and some of sysctls in /proc/sys/kernel and is *bounded*.
>
> Thanks,
> Kirill
>
>
> Michael Portz wrote:
>> Hi!
>>
>> Everywhere it is stressed, that CPUUNITS is only relevant for the
>> relative amount of time a VE has control of the resources. I am
>> looking for a quite different answer: Does it have an absolute
>> meaning as well?
>>
>> E.g. if VE0's CPUUNITS=1000, VE1's CPUUNITS=1000 and VE2's CPUUNITS=1000
>> (and these are *all* VEs), then they all get the same share of processing
>> time. The same holds for CPUUNITS=100 for all VEs. But is there any
>> difference for the values 100 and 1000 respectively?
>>
>> Does e.g CPUUNITS=100 mean, a VE is swapped out after 100 * c timeunits
>> for a constant c? Specifically: Does decreasing the values of CPUUNITS
>> decrease the reaction time of VEs towards interrupts etc.?
>>
>> Thanks
>> Michael
>>
>>
--
accom GmbH & Co. KG
Gr�ner Weg 100
52070 Aachen

Tel: +49 241 918 5228
Fax: +49 241 918 5299

Report message to a moderator

Previous Topic:	Help with OpenVZ forum password reset
Next Topic:	Network slowness

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Wed May 31 10:29:24 GMT 2023

Total time taken to generate the page: 0.01380 seconds