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

Re: [Condor-users] gang scheduling multiple CPUs on an SMP machine



On Tue, 11 Jan 2005 07:50:40 -0500 (EST), Hahn Kim <hgk@xxxxxxxxx> wrote:
> Hello,
> 
> We have a user running a multi-threaded MPI application, i.e. each rank
> itself is multi-threaded.  Our cluster consists of dual-Xeon SMP machines
> and we set NUM_CPUS to 2 in Condor.
> 
> The problem is that the MPI application uses an Intel math library that
> only allows a single process to use the library in a multi-threaded
> manner.  However, Condor often allocates two processors on the same
> machine to two ranks.  When threads from both ranks attempt to access the
> library, the application fails.

The simplest, though not overall best, way to do it is for your end
users to ensure they all add the following to their licence restricted
applications

requirements = VirtualMachineId == X

where X is 1 or 2*

this limits you to only two different sets of applications with
licence restrictions (not to mention that other jobs may run at the
same time which may not give you the best performance)

This is however the fastest work round for such issues and I know it
works just fine without side effects assuming well behaved users (and
if they aren't just restrict their priority so much they only get 1
job at a time :).

*assuming dual CPU, if you have more then adapt as required

Matt