Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
113 lines (67 sloc) 5.7 KB

Parallel MATLAB Server

MATLAB supports parallelization on desktop computers which can be used to increase the speed of analysis drastically. MATLAB also provides the MATLAB Parallel Server (previously the MATLAB Distributed Computing Server) which allows you to write MATLAB code on your local desktop or laptop computer and perform the computation using the CARC high-performance clusters. This QuickByte leads you through the steps needed to set this up. If you run into problems please send an email to and we will be happy to help.

Please ensure you have the MATLAB Parallel Toolbox installed on your local comnputer.

MATLAB Parallel Server Client Configuration

Once MATLAB is installed on your local machine (the MATLAB version on your local machine must match the version on the CARC cluster) click "add-ons" to open Add-on explorer. Search for PBS.

Click on the link "Parallel Computing Toolbox plugin for MATLAB Parallel Server with PBS" (there is a plugin for slurm as well).

Click the "install" button and the plugin will install and a wizard is started.


Wizard1 Choose UNIX in the cluster type


Select no for shared job location.


Enter the address of the cluster you would like to use: for example, or

Enter /users/<your username> for the path to the PBS scripts (remote job storage location) that MATLAB will create on the cluster.

Select unique subfolders.


Select the number of workers and number of threads per worker. This may depend on the program you are running but in general you should have one worker per core on the cluster. For set up and validation leave the number of workers at 1. Leave the threads per worker at 1 unless your software requires more threads.

Specify the path to the matlab installation on the compute nodes: /opt/matlab/R2018b for the Xena cluster, and /opt/local/MATLAB/R2018b for the Wheeler cluster.


Choose flexnet for the license


Name your profile. For example "xena-distributed"


Review your profile settings and create the profile.


You can create multiple profiles for different CARC clusters and numbers of workers.

Validating the Configuration

Select "parallel" then create/manage clusters.

Choose the profile you just created. In this example, the profile name is "xena-distri"


Select the profile you just created and select the validation tab. Press the validate button.


It will ask for your CARC username, and here you can also select your ssh keyfile if you use one (cmd+shift+. to reveal hidden directories so you can see you ~/.ssh folder and select your private key), or just enter your password.

MATLAB will now validate your setup. If you run into trouble please contact CARC support at

This completes the install and configuration.

We have found it best to restart MATLAB at this point, otherwise setting the hostname in the next step may not work.

Writing Parallel Matlab Code

When your program runs you may be prompted to allow incoming/outgoing connections on your computer.

You will likely have to tell the system the IP address or hostname of your local machine. This is so the CARC cluster can communicate with your laptop or desktop. You will set the hostname with pctconfig (Parallel Config Toolbox). You can either type the hostname in directly or attempt to have MATLAB find it for you with the following commands:


[~,name]=system('ipconfig getifaddr en0');


[~,name]=system('hostname -i');

Windows 10

Look up your computers IP address and enter:

pctconfig('hostname',"<your IP address>");

Once you have set the hostname you can run your parallel MATLAB code. Mathworks provides extensive documentation on using parallelism in MATLAB: Mathworks Docs

The very simple program that follows demonstrates how to run parallel code using MATLAB. The code uses parfor, which can often be used to replace a for loop in serial MATLAB code, to distibute the work across 10 parallel workers.

n_workers = 10;                  % We will request 10 workers to run in parallel
p = parpool('xena-distributed', n_workers); % Create the pool of workers using the profile created earlier with 10 workers.
parfor i = 1:100                % Define a parallel loop that will be distributed accross the 10 workers.
i                               % Print the value of i for this iteration.
delete(p);                      % Clean up the worker pool
You can’t perform that action at this time.