GLX Cluster Setup and Usage Instructions

The GLX cluster is available for use by this class:

- A 5-node cluster ("glx nodes") which you will need to access from within the EECS firewall. Each of these nodes contains a quad-core Intel(R) Core(TM) i5-3330 CPU processor running at 3GHz, and 4GB RAM (shared by the cores). The glx nodes are labeled glx1.eecs.wsu.edu through glx5.eecs.wsu.edu. They are conencted by a gigabit ethernet switch. Therefore you can use this system as a "cluster". There is no support for job queue system at this point of time. Instructions on how to run your parallel jobs are specified under Setup.

- It is expected that the students will also have restricted access to another (larger) cluster within VCEA. More details will be posted here as that arrangement is finalized.

The above will be the only dedicated platforms made available for this course. Each student is welcome to use other clusters (for MPI codes) or laptops (for OpenMP/multithreaded programs) that he/she may have access to. If a student uses a different platform then the details of that platform will have to mentioned in the experimental setup section of each programming project report.

Setup:

•  Instructions to set up MPI daemon on the cluster PDF
• Example machine file machinefile.txt
• Commands to run MPI jobs on the cluster:
      > To compile, use the command:      mpicc -o <exec name> <sourcefiles>
      > To run your jobs, use the command:  mpiexec -machinefile <path>/machinefile.txt -n <number of procs> <path>/executable-name <args if any>

    OpenMP instructions:

• To compile your OpenMP function you need to use the -fopenmp flag. For instance to compile a C/OpenMP code on the glx nodes you will use "gcc -fopenmp -o <execname> <sourcefile(s)>".
• To run the executable, simply run the executable directly with its arguments like you would do if it were a serial code. No need to do anything different. It is a good practice to have the user specify the number of threads as one of the executable's arguments.