MATLAB Distributed Computing Server lets you run intensive MATLAB programs and Simulink models on the Tandy Community Supercomputer. You develop your program or model on a desktop computer with Parallel Computing Toolbox, then run the job in a batch with the installed Distributed Computing Server on Tandy. Our license allows running up to 32 simultaneous Matlab workers.
Tandy's license for MDCS covers any academic users only who already have their own (or their institution's) license for Matlab. You can use any toolboxes that you have an Academic license to (Student licenses will allow MATLAB to run on the supercomputer, but not the toolboxes). You need a license for Parallel Computing Toolbox (It appears that a Student license for this will suffice, but not for any other toolbox), but you do not need a license for Matlab Distributed Computing Server - we already have that.
Version Compatibility Note
The version of Matlab connecting to Distributed Computing Server must match exactly with the version of Distributed Computing Server.
Check the Software Resources page to determine which versions are currently supported.
Additional versions can be installed as needed by sending a request to firstname.lastname@example.org.
You will need to configure your local (on your own computer) copy of MATLAB to connect to the supercomputer. You should only need to do this once per computer you use to connect. Do the following:
/home/commonFiles/matlab/RemoteJobSubmissionSetup.zip. You can also download it here: RemoteJobSubmissionSetup.zip.
addpath, documented on the MathWorks site here. This can also be done from the graphical interface.
chmod u+x *.sh
On the supercomputer: Create a MATLAB working directory to hold similar temporary files on the supercomputer side. This can be anywhere that you can access, but generall we suggest creating a folder in your home directory called
matlab-temp. Do this by executing the following command:
setupTandyProfile. It will ask you for the following pieces of information:
Show the list of available cluster profiles with
The Tandy profile name will depend on the version of MATLAB you are using.
Next start a connection to the cluster, using the name of the Tandy profile found above:
You will need to set your code up to use the MATLAB batch job functionality. Information on this functionality is available from the MathWorks documentation for the
batch function. An example execution might look like this (we'll be using the Parallel Computing Toolbox's built-in parfor benchmarking function):
batchfunction for more information.
waitfunction to block until the job is finished. Refer to the MathWorks documentation for the
You can quit your local copy of Matlab while jobs are running. To see a list of your jobs on the cluster from a new Matlab session, do the following:
To remote submit jobs to MATLAB, we use a script provided by MATLAB to mirror job files over SFTP and execution job submission over SSH. This script can go slowly. Please let us know if you need help resolving this issue by contacting email@example.com.