A Brief Introduction to the Springfield Cluster.

• Info

  The Springfield cluster is composed of 24 diskless dual-opteron 1U compute nodes, 1 login master node and 1 4U fileserver.

• Access

  First you must get an account on the cluster from the system administrator. Once you have an account you can access the cluster thru ssh. If you are using Macintosh OS X you can use the teminal program. Linux users can use ssh from the command line, and Windows users can use putty. To run programs on the cluster, which requires to open a window (e.g. emacs) you must login with the -Y option ssh (ssh -Y springfield.physics.ucla.edu). Your computer must also support the Linux window management, which is native on Linux machines. For MacOS X use X11 and Cygwin for Windows.

  For file transfer you can use either scp or sftp. MacOS X and Linux can use the command line scp program. For MacOS X fugu is an available GUI for sftp. Winscp and filezilla are free and easy to use guis for Windows.

• Scientific Codes Installed on Springfield:

  • Tredi - mpi version
  • Genesis
  • QUINDI
  • Spur - mpi version
  • Gdfidl
  • fieldEye
  • OopicPro - mpi version
  • Elegant
  • Parmela

• Guidelines for Using the Cluster.

  • User should always log onto springfield, which redirects to the head node of the cluster. The internal name is homer. Users can log on other nodes but it is strongly discouraged.
  • Jobs should be executed through the Sun Grid Engine. If a user needs an interactive run, he can start an interactive session through the gridengine as well with qsh.
  • Code development and testing can be also done in a limited range on the head node. The main purpose of the head node is running the gridengine and the NFS server for the nodes as well managing the network connection to the outside network.
  • There are two mpi implementation on the cluster, lam-mpi and mpich, but due to its integration with the sun gridengine, it is strongly recommended to use mpich.