NZIAS High Performance Computing Cluster

User Tools

Site Tools



The CTCP Double-Helix Cluster Marvin Node is a compute cluster located on the Albany campus of Massey University. The cluster belongs to the CTCP which is part of the NZIAS.


<WRAP center round todo 60%> todo box </WRAP>

In order to register for use of the cluster an email should be sent to Joshua Bodyfelt. Please include the following details:

  1. Massey University Department,
  2. Massey Permanent Staff Sponsor (i.e. Research Group)
  3. Approximate needed CPU hours
  4. Required Software

Getting Started

The cluster is heavily used by a number of different research groups. In order to avoid impacting other users, all computations should be carried out by submitting jobs to the Torque/PBS queuing system. Please do not run any calculations directly on the login nodes.

A short guide for the Torque/PBS queuing system can be found here.

Example PBS Jobs

We have prepared a number of example PBS job scripts for various programs. You can find them by logging into the cluster and navigating to /data/programs/Example_Jobs. Each folder in this directory contains a job script (with a .sh extension) together with example input and output files. The example located at /data/programs/Example_Jobs/Basic_Job/ provides a simple framework for creating new PBS scripts.

In order to run one of these examples you should:

  • Create a directory in you home folder.
  • Copy the relevant job script to this directory and give it a sensible name.
  • Copy the required input files to the same directory.
  • Edit the job script, as described by the comments in the file.
  • Submit the job to the queue using the 'qsub' command.

Your job will now be listed on the queue, which can be seen by running the qstat command. When a computer becomes available the queuing system will start your job running.

Important Directories

There are three folders where you can place data:

  • /home/$USERNAME
  • /data/$USERNAME
  • /scratch/$PBS_JOBID

The first two directories are available on every login and compute node.

/home/$USERNAME - is your home directory. You can place up to 10GB of data here.

/data/$USERNAME - is on a large 20TB RAID system. There is no hard limit to how much data you can store here. This directory will not be backed up.

/scratch/$PBS_JOBID - is created on the relevant compute node when one of your jobs is executed. It will be immediately deleted once your job has completed. As it is on the HDD of the compute node it will be much faster than the /home/$USERNAME or /data/$USERNAME directories. Therefore you should use it for storing all the temporary files your jobs create. Be careful to copy all important data/results somewhere else before the job finishes and they are deleted.


The cluster and documentation are still being improved. If you have any comments or suggestions please contact either Joshua Bodyfelt or Andrew Punnett.

start.txt ยท Last modified: 2014/10/16 11:13 (external edit)

Page Tools