On Tue, May 30, 2006 at 04:12:55PM +0200, Mehdi Bennani wrote:
> Hello,
>
> I've read the pages about vsched on
>
> http://linux-vserver.org/vsched+explained:
> http://linux-vserver.org/Scheduler+Parameters
>
> I'm trying to figure out how vserver deals with multi-processor
> systems and systems with hyperthreaded processors (detected as
> bi-processor).
>
> and I would like an explanation about something that seems ambigous
> to me:
>
> --Both pages seem to say that (fill-rate) / (interval) is the
> proportion for ONE processor in the system. But a sentence in
> http://linux-vserver.org/vsched+explained...:
>
> "So, if you have four CPUs and you want one context to get an average
> of one whole CPU to itself, then you'd set fill-rate to 1 and interval
> to 4."
>
> ... is not clear... does it mean that (fill-rate) / (interval) is the
> proportion for the FOUR CPU's?
>
> Where is the truth?
as usually the truth is somewhere inbetween ...
actually we have to differentiate here between
the stable and the devel releases, because:
stable has a single token bucket for all cpus
and the scheduler adds 'rate' tokens per cpu
and 'interval'
development has one token bucket per cpu and
two rate/interval values for each of them, to
control normal and idle time
> --Is a processor with Hyperthreading counted as two processors
> by vserver?
yes, as SMT appears quite similar to SMP at
least regarding the scheduler and runqueues
HTH,
Herbert
> Thanks,
>
> Mehdi Bennani
> _______________________________________________
> Vserver mailing list
> Vserver@list.linux-vserver.org
> http://list.linux-vserver.org/mailman/listinfo/vserver
_______________________________________________
Vserver mailing list
Vserver@list.linux-vserver.org
http://list.linux-vserver.org/mailman/listinfo/vserver
Received on Tue May 30 18:19:49 2006