High Performance Computing

  • High Performance Computing

    The advent of high performance commodity processors, plus the emergence of a robust open Unix system (Linux), has made possible the development of inexpensive local clusters of multiple-processor systems for medium-scale computations. Our cluster, MPhase, is designed to serve computational needs of students and faculty at the School of Engineering, Rutgers University.

  • MPhase Overview
    • Hardware

      The mphase cluster consists of over 20 servers, containing dual core processors and operating on a Linux-based operating system. The technical specifications for these machines are listed below:.

         Intel Pentium(R) D CPU @ 3.20GHz (Dual Core, 64-bit)
         2GB DDR2 240-pin unbuffered RAM @ 677 Mhz (PC5300)
        Gigabit (1000MBit/s) Network
    • Software

      The operating system installed is the latest Ubuntu LTS. Cluster usage and management is done through the Oracle Grid Engine (SGE) software queuing system (link found here, please try to find links for the other software packages below and create clickable links to their respective webpages: Oracle Grid Engine Site

      Software installed includes:

         GNU Debian Compiler Tools (C/C++, Java, Fortran, MPI, and many others)
         Matlab (version R2010a)
        Ansys 12

      Additional software may be installed upon request

    • FAQ

      1. How do you access MPhase cluster? Access is provided to the MPhase cluster through secure shell (SSH). X Forwarding is enabled on the system. For example: ssh -X user@mphase.rutgers.edu

      2. How do you access an individual node once logged into MPhase cluster? The command qlogin will provide you an available machine in the cluster with the least resources in use.

      3. How do you submit your job to the queuing system? qsub (scripttorun).sh

      The simplest batch script is batch_run_simple.sh. It can be used for any generic serial job. However if your application writes/reads a lot of data (I/O to files), you'd get much better performance utilizing a local node /tmp directory. Use batch_run.sh submit script in such a case. 

      4. How do you compile a program using gcc compiler tools? gcc -o (outputfile) (inputfile)

      5. How do you execute MPI programs? (ask me personally about this, I think you need to daemonize the mpd program with mpd & first, then you can submit the job but im not sure).

      6. How do you check the status of a job? To check by job ID: qacct -j (Job ID) To check by username: qacct -j -o scarvel

      7. How do you use the graphical interface? qmon (Needs X forwarding enabled in SSH client)

      8. Help! I accidentally deleted a file! No problem, we make regular backups of the files contained in your home directory. Contact us and we may be able to recover your files for you.