shadow_tr

Economics computing at Tulane

  • Advice on statistical software

  • High-performance computing
    • The Department of Economics has dedicated computing resources through the Center for Computational Science (CCS). The CCS cluster has 192 cores with 384GB of RAM, a 40GB/s Infiniband network, and a 19TB scratch disk. Of this, our department has priority on one node with 32 cores and 128GB of RAM, and a 15TB storage array. MATLAB 7.7 is available with a variety of packages, including those for parallel processing. A 10-user, 16-core license of Stata-MP 12 is also installed on the server. Email Keith Finlay if you have any questions.


    • Obtain a CCS account. Mention economics in the request. Let Keith know when you've received confirmation, he'll get you added to the econ workgroup.
    • Using Stata on the CCS server
    • Using Matlab on the CCS server
    • MATLAB packages available:

      • Simulink 7.2
      • Control System Toolbox 8.2
      • Genetic Algorithm and Direct Search Toolbox 2.4
      • Image Processing Toolbox 6.2
      • MATLAB Compiler 4.9
      • Mapping Toolbox 2.7.1
      • Neural Network Toolbox 6.0.1
      • Optimization Toolbox 4.1
      • Parallel Computing Toolbox 4.0
      • Signal Processing Toolbox 6.10
      • SimPowerSystems 5.0
      • Simulink Control Design 2.4
      • Statistics Toolbox 7.0
      • Symbolic Math Toolbox 5.1
      • Wavelet Toolbox 4.3
    • Using R on the CCS server
    • Using Fortran on the CCS server
    • Using the CCS server to backup data
    • Queues available on Sphynx
    • Queue Name

      Max Walltime

      Memory

      Max # Cores / User

      Priority

      Nodes in Queue

      Core / Nodes

      Total Cores in Queue

      econ_long

      300 hrs

      32 GB / core

      64

      High

      1

      32

      128

      ccs_short

      6 hrs

      2 GB / core

      128

      High

      7

      32

      224

      ccs_long

      168 hrs

      2 GB / core

      64

      Low

      4

      32

      128

      Walltime is human time, regardless of CPU time. The total number of cores available to a user is 128 regardless of the queues in use. This number will increase as more nodes are added to the cluster. User job priority is a function of time-decayed use in the last seven days.

206 Tilton Hall, Tulane University, New Orleans, LA 70118 tel 504-865-5321 fax 504-865-5869 pwatson@tulane.edu