Parallel Evaluation CONntroller:
Run Matlab in Parallel on your Linux/Unix Cluster

PECON (Parallel Evaluation CONtroller) is a very small, easy-to-use Matlab class library that simplifies the task of parallelizing existing Matlab programs. PECON exploits Matlab's Java Virtual Machine to pass XML-encoded data structures between a central client and several "compute servers", thereby avoiding reliance on lower-level message-passing software. This arrangement is ideal for "embarrassingly parallel" applications in which time is mainly spent on computing results from small amounts of data. PECON works on a multiprocessor machine or network running Unix, such as a Beowulf cluster, standard Linux cluster, or multiprocessor Macintosh. All processors must have access to the same file system. You must have version 6 or later of Matlab installed on this file system, and socket calls must be enabled. You must be able to ssh to all the hosts that you wish to use, without having to enter a password. In order to do this, run the following commands from your home directory: 
  % ssh-keygen -trsa 
  (hit return to use defaults; including no passphrase)
  % cd .ssh
  % cp id_rsa.pub authorized_keys
To try out PECON, download and unzip the file pecon.zip. Then download and unzip the XML Toolbox. Change startup.m to reflect where you put the resulting files, and make sure to copy startup.m to any directory from which you want to run PECON. To run PECON, you'll need to know the names of some nodes to which you can ssh from the node on which you're running (because PECON automatically calls ssh to run Matlab remotely on the nodes). Let's say you have five such nodes, named alpha, bravo, charlie, delta, and echo, and that your sysadmin prefers you to use port 15000 for personal projects. You can then issue the following instructions in Matlab. 
  >> p = pecon({'alpha', 'bravo', 'charlie', 'delta', 'echo'}, 'port', 15000)
After a brief startup period, Matlab should respond with 
  Server alpha: ready
  Server bravo: ready
  Server charlie: ready
  Server delta: ready
  Server echo: ready
You can then try (trivial) parallel evaluation of a function by doing 
  >> feval(p, @sqrt, {0, 1, 4, 9, 16, 25, 36, 49, 64, 81})
Matlab should respond with 
  ans = 

    [0]
    [1]
    [2]
    [3]
    [4]
    [5]
    [6]
    [7]
    [8]
    [9]
You can shut down the servers by 
  >> halt(p)
Now you are ready to use PECON to parallelize existing Matlab functions on your Linux/Unix system. The utils directory contains a few utility functions for killing errant servers, and partitioning and re-assembling matrices. The technical paper describing PECON is available here. Please e-mail levys@wlu.edu with any question or complaints. Enjoy!