"round" is an OpenMPI ring ping to demonstrate a message getting passed between each node and modified one at a time

round.c

#include <stdio.h>
#include <string.h>
#include <unistd.h>
#include <mpi.h>

#define MASTER  0
#define TAG     0
#define MSGSIZE 100000
#define MAX 25

int main(int argc, char* argv[])
{
    int my_rank, num_nodes, last_node;
    char my_host[MAX];
    char message[MSGSIZE];

    MPI_Init(&argc, &argv);
    MPI_Comm_rank(MPI_COMM_WORLD, &my_rank);
    MPI_Comm_size(MPI_COMM_WORLD, &num_nodes);
    last_node = num_nodes-1;
    gethostname (my_host, MAX);

    if (my_rank == MASTER) {
        printf("there are %d nodes for this job\n", num_nodes);
        sprintf(message, "this note has been seen by: %s(%d)", my_host, my_rank);
        MPI_Send(message, strlen(message) + 1, MPI_CHAR, 1, TAG, MPI_COMM_WORLD);
        // get "signed note" back from the last node
        MPI_Recv(message, MSGSIZE, MPI_CHAR, last_node, TAG, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
        printf("%s\n", message);
    } else {
        MPI_Recv(message, MSGSIZE, MPI_CHAR, my_rank-1, TAG, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
        sprintf(message, "%s, %s(%d)", message, my_host, my_rank);
        if (my_rank < last_node) {
            // this is a node (and not the last node)
            MPI_Send(message, strlen(message) + 1, MPI_CHAR, my_rank+1, TAG, MPI_COMM_WORLD);
        } else {
            // this is the last node; send message to master
            MPI_Send(message, strlen(message) + 1, MPI_CHAR, MASTER, TAG, MPI_COMM_WORLD);
        }
    }

    MPI_Finalize();

    return 0;
}

Compile with mpicc round.c -o round. Run with mpirun -np 80 --hostfile ../my_hosts round. You will need to add an --oversubscribe flag or reduce the nodes if you don't have 80. The hostfile flag should point to an actual file with lines like machine01 slots=8 where the first part of each line is a machine name and the second is the number of threads to use (typically you can use 2 threads per core, but you might want to leave room for other jobs). You should also make sure your user is able to run ssh commands on the other nodes via proper ssh configuration.