Designing MPICH-Cluster

 

MPICH is a high performance and widely portable implementation of the Message Passing Interface (MPI) standard. 

 

What it does:

1. Provides high computing interface for conducting cutting edge research in MPI.

2. Enables memory share amoong multimple computing platforms.

3. Supports high-speed networks and proprietary high-end computing systems.

​

I personally love playing around with Ubuntu Server. In this tutorial we will use 3 nodes with the host name love0, love1 and love2 running on Ubuntu server 12.04 LTS edition. 

 

Step 1: Edit /etc/hosts file as shown below

 

127.0.0.1    localhost

192.168.1.10    love0

192.168.1.11    love1

192.168.1.12    love2

 

Step 2: Install NFS

 

ambuj@love0:~$ sudo apt-get install nfs-server

 

Step 3: Making and sharing the main folder

 

ambuj@love0:~$ sudo mkdir /mirror

 

edit /etc/exports file using-

ambuj@love0:~$ echo "/mirror *(rw,sync)" | sudo tee -a /etc/exports

 

now restart the nfs-server

ambuj@love0:~$ sudo service nfs-kernel-server restart

 

Step 4: Mount master node from all other nodes

ambuj@love1:~$ sudo mount ub0:/mirror /mirror

ambuj@love2:~$ sudo mount ub0:/mirror /mirror

 

or we can edit /etc/fstab file and add the below shown line in order to mount the master in in every boot

love0:/mirror    /mirror    nfs

then remount all partitions from other nodes using

ambuj@love1:~$ sudo mount -a

ambuj@love2:~$ sudo mount -a

 

Step 5: Define a user with same name and same userid in all nodes with a home directory in /mirror

Here we use the name "heart"

ambuj@love0:~$ sudo adduser heart

add user as sudoer by

ambuj@love0:~$ sudo adduser heart sudo

 

Now change the owner of /mirror to heart

ambuj@love0:~$ sudo chown heart /mirror

 

Step 6: Install SSH Server

ambuj@love0:~$ sudo apt­-get install openssh-server

 

Step 7: Setting up passwordless SSH for communication between nodes

 

First we login with our new user to the master node:

ambuj@love0:~$ su - heart

 

Then we generate an RSA key pair for heart:

heart@love0:~$ ssh­-keygen ­-t rsa

You can keep the default ~/.ssh/id_rsa location. It is suggested to enter a strong passphrase for security reasons.

 

Next, we add this key to authorized keys:
heart@love0:~$ cd .ssh
heart@love0:~/.ssh$ cat id_pub.dsa >> authorized_keys

As the home directory of heart in all nodes is the same (/mirror/heart) , there is no need to run these commands on all nodes. If you didn't mirror the home directory, though, you can use ssh-copy-id <hostname> to copy a public key to another machine's authorized_keys file safely.

 

To test SSH run:
heart@love0:~$ ssh love1
or
heart@love0:~$ ssh love2

 

If you are asked to enter a passphrase every time, you need to set up a keychain. This is done easily by installing... Keychain.
heart@love0:~$ ssh love1~$ sudo apt-get install keychain

And to tell it where your keys are and to start an ssh-agent automatically edit your ~/.bashrc file to contain the following lines (where id_rsa is the name of your private key file):
if type keychain >/dev/null 2>/dev/null; then
  keychain --nogui -q id_rsa
  [ -f ~/.keychain/${HOSTNAME}-sh ] && . ~/.keychain/${HOSTNAME}-sh
  [ -f ~/.keychain/${HOSTNAME}-sh-gpg ] && . ~/.keychain/${HOSTNAME}-sh-gpg

 

Exit and login once again or do a source ~/.bashrc for the changes to take effect.
Now your hostname via ssh command should return the other node's hostname without asking for a password or a passphrase. Check that this works for all the slave nodes.

 

If it still doesnt work then try following:

Copy public key to a remote host:

heart@love0:~$ ssh-copy-id -i ~/.ssh/id_rsa.pub ambuj@love1                 #this

 

heart@love0:~$ cat ~/.ssh/id_rsa.pub | ssh ambuj@love1 \
      'mkdir -p ~/.ssh ; cat >> ~/.ssh/authorized_keys'                               # or this

 

Step 8: Install GCC

heart@love0:~$ sudo apt-get install build-essential

 

Step 9: Install MPICH2

ambuj@love0:~$ sudo apt-get install mpich2

 

Step 10: Setting up a machinefile
Create a file called "machinefile" in heart's home directory with node names followed by a colon and a number of processes to spawn:

 

love2:2  # this will spawn 2 processes on love2
love1    # this will spawn 1 process on love1
love0    # this will spawn 1 process on love0

 

Step 11: Testing

 

Change directory to your mirror folder and write this MPI helloworld program in a file hello.c (courtesy Ubuntu blog):

 

#include <stdio.h>

#include <mpi.h>

 

int main(int argc, char** argv) {

    int myrank, nprocs;

 

    MPI_Init(&argc, &argv);

    MPI_Comm_size(MPI_COMM_WORLD, &nprocs);

    MPI_Comm_rank(MPI_COMM_WORLD, &myrank);

 

    printf("Hello from processor %d of %d\n", myrank, nprocs);

    MPI_Finalize();

 

    return 0;

}

 

Compile it:

heart@love0:~$ mpicc hello.c -o hello

 

and run it (the parameter next to -n specifies the number of processes to spawn and distribute among nodes):

 

heart@love0:~$ mpiexec -n 4 -f machinefile ./hello

 

You should now see output similar to this:

 

Hello from processor 0 of 4

Hello from processor 1 of 4

Hello from processor 2 of 4

Hello from processor 3 of 4

 

Enjoy!!!!!!!!! Your MPI Platform is ready