JCU HPC provides a few options for running interactive jobs:

  1. Windows session hosts - for software only available on a Microsoft Windows operating system.
    1. You will need to submit a request for access to Windows session hosts.  One or meetings are likely to be arranged to identify requirements.
    2. Local storage on the session hosts operates as "scratch space" - no protection, scheduled deletion of old files.
    3. The amount of CPU and memory resource available on session hosts is small.
  2. HPC cluster login nodes - for development and testing or workloads that are not resource intensive.
    1. If your work is GUI based, you will need to use a connection that can pass graphics back to your personal computing device.
    2. There are only two login nodes, each with 40 cores (80 threads) and 384GB of memory.
  3. HPC cluster compute nodes - for any workload that can be completed on a compute node.
    1. Idle interactive jobs may be deleted by HPC staff during times of high-resource demand.
    2. You will only get a prompt (logged into a compute node) back quickly if sufficient free resources exist to be dedicated to your work.
    3. For command line work, your job submission command might look like: 

      qsub -I -l select=1:ncpus=1 -l mem=4gb -l walltime=2:00:00

    4. For GUI work,  your job submission command might look like: 

      qsub -X -I -l select=1:ncpus=1 -l mem=8gb -l walltime=1:00:00

Note:  Many HPC facilities do not allow running of interactive work on a cluster, due to the amount of resource wastage involved in interactive work.  Some may provide virtualized platforms for interactive work - similar to what JCU is providing for computational work performed on a Microsoft Windows operating system.  JCU HPC is likely to return to virtualized login nodes in future.