+Open all         -Close all

• About DCSC/SDU
  • Home
  • Board
  • News
• Resources
  • Common infrastructure
  • Cluster: Horseshoe6
  • Cluster: Horseshoe5
  • Cluster: Horseshoe4
• Statistics
  • Utilisation Graphs
• Support
  • How to Get a Grant
  • How to Get an Account
  • Docs
    • FAQ
    • Intel MKL
    • MAUI
    • PETSC
    • PGI
• Contact
  • DCSC/SDU Contact
  • DCSC Central Office
  • Other DCSC Centers

 Dear User,

 Welcome to DCSC's Super Computer at SDU (aka. The Horseshoe).

 We have prepared some documentation for how to use the facility.
 The following questions are answered:

 -- what kind of computer is the The Horseshoe ?

        Please read "horseshoe"
 
 -- where's my diskspace ?
 -- is it backed up ?     

        Please read "diskspace"

 -- how do I change password ?
 -- how do I change my shell ?

        Please read "password"

 -- what is PBS/TORQUE ?
 -- how do I run a job ?
 -- how do I see which jobs I'm running ?
 -- how do I kill a running job ?

        Please read "pbs" and "MAUI"

 -- which PBS queues are available ?
 -- what are the resource limits on jobs in the different queues ?

        Please read "PBS queues"

 -- how can the new multi-core nodes be used more efficiently ?

        Please read "Jobs on multi-core machines"

 -- is there something I can do to make my job(s) start sooner ?

        Please read "Job resource request management"

 -- is it possible to run jobs using more time than the queue limits ?

        Please read "PBS multipart jobs"

 -- how can I avoid loosing files because the job walltime expires ?

        Please look at this example PBS script: pbs_timer_job_script

 -- can I avoid that a PBS job is requeued after an initial failed run ?

        Yes, add '-r n' to the 'qsub' command or '#PBS -r n' in your jobscript.

 -- how do I compile my programs ?
 -- how do I compile my programs with large file support (files > 2GB) ?
 
        Please read "compile"

 -- how do I compile and run MPI programs ?

        Please read "OPEN-MPI"

 -- how do I compile and run PVM programs ?

        Please read "PVM"

 -- how can I make sure my job uses the 8GB nodes ?
 -- how can I control which switches (groups of nodes) will be used for my job ?

        Please read "Controlling switch-fragmentation

 -- where do I find information about:
      -> usage on the cluster right now ?
      -> loadlevel on nodes right now ?
      -> my past usage on the cluster ?
      -> the queue right now (sorted by start-priority) ?

        Please look at "Cluster Statistics"

 -- why is The Horseshoe utilization not 100% - the queue is never empty !? 

        Please read "Horseshoe Utilization"

 If you have questions not answered here please send e-mail to
 admin@dcsc.sdu.dk