== Cluster Controller Usage Tutorial == == Introduction == This tutorial guides the user through an example of using the cluster worker controller component of CML. The tutorial is completed entirely on the cluster, so the only software requirement is that the user have an SSH client. A full description of the cluster controller software is given in the file ./CMLROOT/doc/desc.txt == Begin Tutorial == Notation: > denotes a command executed on the cluster head Linux command line >> denotes a command executed within MATLAB on the cluster head 1. Log in to cluster using SSH. URL: http://wcrlcluster.csee.wvu.edu 2. Create a directory to house the EE 561 project code, and switch to this directory. > mkdir ee561/ > cd ee561/ 3. Download the EE 561 project code and decompress. > wget http://iscml.googlecode.com/files/cml_ee561_v3.zip > unzip cml_ee561_v3.zip 4. Start MATLAB > matlab 5. Initialize CML. >> CmlStartup 6. Create a cluster controller object. The configuration file used in this example utilizes 8 workers on node01 12 workers on node11 >> cwc_obj = cwc(cml_home, 'test.cfg', 'stub_worker') 7. Start 8 workers on node01, and 12 workers on node11. >> cwc_obj.cSta() 8. Observe the 8 worker processes on node01 >> !ssh node01 'ps aux|grep -i matlab' 9. Observe the 12 worker processes on node11 >> !ssh node11 'ps aux|grep -i matlab' 10. Stop the worker processes. >> cwc_obj.cSto() 11. Observe that all worker processes on node01 and node11 have stopped. >> !ssh node01 'ps aux|grep -i matlab' >> !ssh node11 'ps aux|grep -i matlab'