Re: [Condor-users] Windows 2000 Server + Condor

["Andrey Kaliazin" <A.Kaliazin@xxxxxxxxxxx>]

Hi Matt

It was not my intention to sound all-knowing and overconfident.
What I meant was that in the particular case of running windows server
on a P4HT hardware the overall system performance (responsiveness) 
is likely to be higher with hyperthreading turned on.
Condor apps may not see the boost, but the OS itself and other, 
non-condor services may.

Also we had a case when on the Xeon-based workstations running Linux
with SMP-aware kernel condor jobs were failing erroneously, when HT
was disabled. With HT on, everything works flawlessly with one 
condor VM per node.

kind regards,

Andrey 


> -----Original Message-----
> From: condor-users-bounces@xxxxxxxxxxx 
> [mailto:condor-users-bounces@xxxxxxxxxxx] On Behalf Of Matt Hope
> Sent: Tuesday, April 24, 2007 1:34 PM
> To: Condor-Users Mail List
> Subject: Re: [Condor-users] Windows 2000 Server + Condor
> 
> On 4/24/07, Andrey Kaliazin <A.Kaliazin@xxxxxxxxxxx> wrote:
> > Hi Andrew
> >
> > I would not recommend you to turn off hyperthreading on P4 machine -
> > you are loosing on performance and multithreading 
> capabilities for nothing.
> >
> > Much easier to put this line into your condor_config.local file -
> >
> > COUNT_HYPERTHREAD_CPUS  = FALSE
> >
> > and restart the condor service.
> 
> Couple of things one this.
> 
> First: Benchmark your apps with/without hyper threading, make sure you
> cover the case where 1 instance per core is running, one instance per
> socket is running and one instance only is running. See which is
> better, see whether this appears to be memory/cache/disk/network bound
> anyway. Measure, measure, measure.
> Any advice stating flat out one or the other is bogus. All our stuff
> ran better without hyper threading on win 2003 server (only allowing
> as many VM's as there were real cores in both cases) so we disable it
> in the bios to make the (OS's) scheduler's job easier. but our tasks
> tend to suck 100% CPU at all times and have very low cache miss rates.
> Note that Hyper-threading implementation changes were made part way
> through the Netburst architecture's life cycle and so you may see
> different behaviour on different machines.
> 
> Second is that COUNT_HYPERTHREADED_CPUS can have issues in some
> versions with some operating systems. If it does and you farm is
> homogeneous or easy to set up you can explicitly state the number of
> VM's you want with NUM_VIRTUAL_MACHINES
> 
> Matt
> _______________________________________________
> Condor-users mailing list
> To unsubscribe, send a message to 
> condor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/condor-users
> 
> The archives can be found at either
> https://lists.cs.wisc.edu/archive/condor-users/
> http://www.opencondor.org/spaces/viewmailarchive.action?key=CONDOR
>