[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Condor-users] Condor on Windows



Dear Liu,
I run long term (up to 100s days) jobs on WinXP pro/Condor pool that is
available at off-peak time.

My code/program supports check pointing natively, so we use recursive DAG to
resubmit the jobs, obviously your code needs to be "aware" of the
environment, such as time, and kill itself safely when needed to initiate a
transfer back to the submit host, and using recursive DAG submit itself to
restart on a new run.

As per compiler, simply use a statically compiled binary or submit the
runtime libs/dlls as input!

I use the commercially available Intel FORTRAN compiler as we only buy Intel
CPUs, which generally gives about 50% - 60% performance benefit over the GNU
one. 
Hope it helps
________________________________________________
Alan Arokiam,
The Materials Modelling Group,
Materials Science and Engineering,
Department of Engineering,
The University of Liverpool,
Brownlow Hill,
Liverpool,
UK.
L69 3GH
 
Tel: 44-(0)151-794-4671
 
www: http://pcwww.liv.ac.uk/~alanca/
 
> -----Original Message-----
> From: condor-users-bounces@xxxxxxxxxxx [mailto:condor-users-
> bounces@xxxxxxxxxxx] On Behalf Of Honggao Liu
> Sent: 06 April 2005 10:38 PM
> To: Condor-Users Mail List
> Subject: Re: [Condor-users] Condor on Windows
> 
> Codor is designed to harness the wasted CPUs. if the machine is idle,
> Condor will run job there and can move job to another idle machine if
> the current machine is not available(the owner come back and begin to
> use the desktop, for example). On windows, Condor doen't support
> checkpoint. If the job has to move, it has to be run from beginning.
> That is my concern.
> 
> On windows machine, we usually don't have shared Filesystem and we
> cannot use condor standard universe. To run C or Fortran code, we need
> run time libaries and soem DLLs.
> 
> On Apr 6, 2005 4:23 PM, Alain Roy <roy@xxxxxxxxxxx> wrote:
> >
> > >1. On windows platform, Condor doesn't support checkpointing, thus it
> > >cannot checkpoint the job and move the job to a different windows
> > >machine if the current machine is not available. If Condor migrates
> > >the job to new windows machine, the job has to run from beginning. How
> > >can a long job be
> > >finished in a Condor pool of windows Desktop computers? (Of coourse,
> > >we can configure not to vacate job, but that is not I want)
> >
> > I don't understand this question, can you clarify it? Condor doesn't
> > migrate jobs at whim, so how long a job runs depends on your policy and,
> > assuming you run jobs on desktops on which user activity removes jobs,
> when
> > your users use their computers.
> >
> > >2.What kind of compilers (versions?) are supported under Condor
> > >Version 6.7.5 in Windows XP Professional platform? For GNU compilers,
> > >does I have to use Cygwin to compile C and Fortran codes?
> >
> > It doesn't matter on Windows. It only matters for standard universe on
> > platforms that support standard universe.
> >
> > Condor will even run your job if you write it in COBOL.
> >
> > -alain
> >
> > _______________________________________________
> > Condor-users mailing list
> > Condor-users@xxxxxxxxxxx
> > https://lists.cs.wisc.edu/mailman/listinfo/condor-users
> >
> 
> 
> --
> Honggao Liu, Ph.D
> High Performance Computing
> Office of Computing Services
> Louisiana State University
> Tel: (225) 578-0235
> Fax: (225) 578-6400
> E-mail: honggao@xxxxxxx
>             honggao.liu@xxxxxxxxx
> _______________________________________________
> Condor-users mailing list
> Condor-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/condor-users