Deploying an MPI cluster using vcluster

This tutorial describes how to use vcluster, a set of scripts that automates the deployment of an MPI cluster on the peak-two cloud.

1. First, connect to the cloud controller using ssh. The IP address is 10.0.3.82. Use vcluster for both username and password.




2. Once logged on to the cloud controller, start vcluster. Specify the name of the cluster and the number of slave nodes to start. In our sample scenario we are going to name the cluster bioinformatics (first parameter) and start 3 (second parameter) slave nodes. The script will run and will take sometime to finish.



The machines/instances used in the cluster as shown in the dashboard.


3. You can now connect to the front end node using ssh and mpiuser as username and password. You will observe that the hostname of the master node is the name you specified when you invoked vcluster with the added -master, thus bioinformatics-master.


4. You can test for connectivity by ping-ing bioinformatics-slave-1, bioinformatics-slave-2, bioinformatics-slave-3. as shown.


5. Compile and run a sample mpi program.




Resources:





Comments