Run MPI On Grid'5000: Difference between revisions

From Grid5000
Jump to navigation Jump to search
Line 71: Line 71:
= Setting up and starting OpenMPI on a default environment using <code class=command>oarsh</code> =
= Setting up and starting OpenMPI on a default environment using <code class=command>oarsh</code> =
* oarsub -I -l nodes=3
* oarsub -I -l nodes=3
<code class=command>oarsh</code> is the default connector used when you reserve a node. To be able to use this connector, you need to specify it on the <code class=command>mpdboot</code> command line. So you can use exactly the same script than for the previous allow_classic_ssh part and replacing <code class=env>--rsh=ssh</code> by <code class=env>--rsh=oarsh</code>.
<code class=command>oarsh</code> is the default connector used when you reserve a node. To be able to use this connector, you need to add the option <code class=command>--mca  plm_rsh_agent "oarsh"</code> to mpirun.


= Setting up and starting OpenMPI on a kadeploy image =
= Setting up and starting OpenMPI on a kadeploy image =

Revision as of 18:02, 2 February 2010

Warning.png Warning

Practical session under construction.

Running MPI on Grid'5000

When attempting to run MPI on Grid'5000 you'll be faced with a number of challenges, ranging from classical setup problems for MPI software to problems specific to Grid'5000. This practical session aims at driving you through the most common uses cases, which are

  • setting up and starting openMPI on a default environment using allow_classic_ssh
  • setting up and starting openMPI on a default environment using oarsh
  • setting up and starting openMPI on a kadeploy image
  • setting up and starting openMPI to use high performance interconnect

Pre-requisite

Overwiew

Currently, the default environment is not the same on every sites, therefore, you don't have the same version of openmpi on every site. If you want to use OpenMPI for a grid experiment, you will have to install your own MPI version; You have two possibility:

  • install Openmpi on your home dir (but you should recompile and install on a all the sites you want to use! )
  • use the same kadeploy image and deploy it on the the sites you want to use

If you are only interested on a single site experiement, you may use the version providede by the default environment.

Compilation

  • Make a reservation (or connect to the compilation machine if available (compil.<node>.site.grid5000.fr) and compile it with the option you like.
mkdir -p $HOME/src/mpi
oarsub -I  
./configure --prefix=$HOME/openmpi/  --with-memory-manager=none
make -j4
  • Install it on your home directory
make install

Setting up and starting OpenMPI on a default environment using allow_classic_ssh

  • oarsub -I -t allow_classic_ssh -l nodes=3
  • Code to test :
$ mkdir -p $HOME/src/mpi
$ vi $HOME/src/mpi/tp.c
#include <stdio.h>
#include <mpi.h>
#include <time.h> /* for the work function only */

int main (int argc, char *argv []) {
       char hostname[257];
       int size, rank;
       int i, pid;
       int bcast_value = 1;

       gethostname (hostname, sizeof hostname);
       MPI_Init (&argc, &argv);
       MPI_Comm_rank (MPI_COMM_WORLD, &rank);
       MPI_Comm_size (MPI_COMM_WORLD, &size);
       if (!rank) {
            bcast_value = 42;
       }
       MPI_Bcast (&bcast_value,1 ,MPI_INT, 0, MPI_COMM_WORLD );
       printf("%s\t- %d - %d - %d\n", hostname, rank, size, bcast_value);
       fflush(stdout);

       MPI_Barrier (MPI_COMM_WORLD);
       MPI_Finalize ();
       return 0;
}


  • Compile your code
$ $HOME/openmpi/bin/mpicc src/mpi/tp.c -o src/mpi/tp
  • Use this script to launch
$HOME/openmpi/bin/mpirun -machinefile $OAR_NODEFILE $HOME/src/mpi/tp

Setting up and starting OpenMPI on a default environment using oarsh

  • oarsub -I -l nodes=3

oarsh is the default connector used when you reserve a node. To be able to use this connector, you need to add the option --mca plm_rsh_agent "oarsh" to mpirun.

Setting up and starting OpenMPI on a kadeploy image

Building a kadeploy image

Note.png Note

You can skip this section and use directly the environment squeeze-ompi

We will create a kadeploy image based on an existing one.

oarsub -I -t deploy -l nodes=1,walltime=2
kadeploy3 -f $OAR_NODEFILE -e lenny-x64-base -k

Then connect onthe deployed node as root, and install openmpi:

ssh root@<node>
cd /tmp/
tar jvxf ~/softs/openmpi-1.4.1.tar.bz2
cd openmpi-1.4.1
./configure --libdir=/usr/local/lib64 --with-memory-manager=none
make -j4
make install
apt-get -y install libblas-dev
tgz-g5k /dev/shm/image.tgz

On the frontend:

scp node:/dev/shm/image.tgz kaopenmpi.tgz

Using a kadeploy image

Setting up and starting OpenMPI to use high performance interconnect

Myrinet hardware : MPICH-MX

  • oarsub -I -l nodes=3 -p "myri2g='YES'"

or

  • oarsub -I -l nodes=3 -p "myri10g='YES'"

Infiniband hardware : MVAPICH

  • oarsub -I -l nodes=3 -p "ib10g='YES'"

or

  • oarsub -I -l nodes=3 -p "ib20g='YES'"

Setting up openMPI to accept private networks as routable between hosts

OpenMpi site : OpenMPI

  • Connect on one site and reserve a node
ssh nancy.grid5000.fr
oarsub -I
  • Connect to another site and reserve a node
ssh rennes.grid5000.fr
oarsub -I
  • Try to launch the code that you previously create between the two reserved nodes