[edit] SSH Password-less with Putty

First, we will install putty on your windows machine.

At www.putty.nl you can find putty.exe, pscp.exe, puttygen.exe, plink.exe, psftp.exe, and pageant.exe; all free software. Download these and put them in the directory C:\Program Files\Putty

Then putty has to be added to the system path. Right click on my computer and select properties.

Select the Advanced tab then click Environment Variables.

Under system variables, select path and click edit. Add the directory C:\Program Files\Putty to the system variables.

The executable puttygen.exe is used to create an SSH-2 DSA key. This will involve creating a public key and a private key, keeping both on your windows machine, and adding the private key to your accepted keys on the cluster.

Run puttygen.exe and generate your DSA key. Change the key comment to matlab_dce-key and save the public key as matlab_dce.pub and the private key as matlab_dce.ppk. Save these both in the putty directory.

Now transfer the public key (matlab_dce.pub) to your home directory on the cluster. You can SSH then SSH file transfer or use another file transfer program such as filezilla (http://filezilla-project.org/download.php).

Put the public key in /home/yourID/.ssh then add it to your authorized_key file.

   cd ~/.ssh
   ssh-keygen -i -f matlab_dce.pub >> ~/.ssh/authorized_keys
   chmod 600 ~/.ssh/authorized_keys

Now we will point matlab to these keys so it can automatically connect to the cluster. Run putty.exe and create a session for hpcc-master. Type hpcc-master.case.edu into the hostname field and use port 22 then save this session under the name hpcc-master.case.edu.

Under the category: connection > SSH > auth point the private key to the one you saved to your putty directory.

Under the category: connection > Data cahnge the auto loging username to your cluster userID.

Now return to Category: session and save this session. Double clicking this session should confirm that you can log in automatically.

[edit] Importing Matlab Configuration

The remote configuration file can be found for download at http://hpcc1-mgmtnode2.case.edu/matlab/2009aRemoteConfig.mat. Download this file to your windows machine.

Run Matlab 2009a. Go to the Parallel tab and select manage configurations.

Select file > import and import the 2009a Remote Config file you downloaded.

Before jobs can be run, some fields must be edited in this file. For the fields: ParallelSubmitFcn and SubmitFcn you should see a capitalized directory. This should be edited to your matlab working directory on the cluster, for example /home/jxm343/matlab.

Save these changes and click start verification. It should show that Resources are found successfully and distributed jobs are tested successfully. (in the image below; I test on my already modified configuration, not the 2009aRemoteConfig)

Currently, our configuration only supports running distributed jobs remotely to the cluster. Parallel tasks are currently not supported. A solution is currently being investigated. A parallel job can be run on the cluster, but it must be run from the client installation on the master node. For more information on this, see the previous section.