AMBER Archive (2009)

Subject: Re: [AMBER] AMBER on a Mac Mini cluster PoC

From: Alan (alanwilter_at_gmail.com)
Date: Mon Aug 17 2009 - 05:50:30 CDT

Hi Abdul,
Don't get confused. Amber is not ready for Xgrid directly as far as I know.
What used to happen is that Openmpi has a implementation suport for Xgrid
and it used to work with openmpi 1.2.x but now it's broken with 1.3.x. See:
https://svn.open-mpi.org/trac/ompi/ticket/1777

Xgrid in openmpi is all about using your xgrid configuration to submit your
jobs from a openmpi application like e.g. pmemd as you'd do anyway. The
advantage I see is that is quite simple and easy to setup a Xgrid cluster
than go through lists os hosts, modifying file (e.g. /etc/hosts), ssh
permission, etc.

But as Ross emphasised, if you running on ethernet, there's few situation
that you can benefit from multi-node cluster and Xgrid won't do anything
better.

But I always wanted to connect 2 mac in network via firewire... It may not
claims a bandwidth as large as gigabit does but I am quite sure that latency
and reliability is much better. Anyway, I have no time do play that now but
maybe you...

Cheers,
Alan

On Mon, Aug 17, 2009 at 09:13, Abdul Rehman Gani <amber_at_infostream.co.za>wrote:

> Ross Walker wrote:
> > Hi Abdul,
>
> Hi Ross,
> >
> > This looks to be an issue with your MPI installation rather than a
> problem
> > with AMBER per se. Have you tested your openMPI installation? Often such
> > things come with test suites or you can download them to test things out.
>
> I have not run a test suite, but a command like:-
>
> mpirun -machinefile machinefile -np 4 ls -l
>
> does return expected results, as does using 'env' or 'hostname' as the
> command.
>
> >
> > Also I note that the error message, suggests TCP connections which to me
> > implies you are using ethernet. Is this correct? If so then you are
> unlikely
>
> Yes I am using gigabit ethernet.
>
> > to see any benefit running in parallel at least for regular MD runs.
> > Ensemble runs such as replica exchange might work. Unfortunately these
> days
> > ethernet is too slow to be of much help in parallel. Thus you could try
> > running in parallel with 2 threads on one machine. I.e. just one entry in
> > your machine file and -np 2. Then run the tests and see how they do. This
> > will at least test your actual parallel installation.
>
> I just did this and it worked fine. Using only the local host's IP in
> machinfile and -np 2, the issuing 'make test.parallel' ran to completion and
> each test either passed or was skipped (required mpi thread to be a multiple
> of 4)
>
> >
> > Then you can try running some simulations yourself (or perhaps using the
> > benchmarks here http://ambermd.org/amber10.bench1.html) across both
> nodes so
> > you can compare performance. My guess is that it will probably actually
> be
> > slower using 2 nodes rather than 1 but you might be lucky. It will
> certainly
> > be easier to debug MPI problems (often caused by flakey interconnect
> > hardware) when running a job yourself rather than inside the test
> > environment.
>
> Will try that. Ignoring for the moment that I am using Ethernet and the
> provisos you highlighted, how would I use Amber over XGrid instead of
> OpenMPI?
>
> Abdul
>
> >
> > Good luck,
> > Ross
>
>
> _______________________________________________
> AMBER mailing list
> AMBER_at_ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
> 

-- 
Alan Wilter S. da Silva, D.Sc. - CCPN Research Associate
Department of Biochemistry, University of Cambridge.
80 Tennis Court Road, Cambridge CB2 1GA, UK.
>>http://www.bio.cam.ac.uk/~awd28<<
_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber