Child pages
  • Why use JCU HPC?
Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 9 Next »

JCU HPC provides entry-level research computing resources.  Other organisations (QCIF, NCI, Pawsey, etc.) provide much larger, high-capability resources.

Storage

  • You want/need to be doing responsible research.
  • Your research data is too large to store on a personal computing device.
  • Data protection - disk failures are unlikely to result in data loss.
  • Protection against silent data corruption.
  • Proximity to HPC cluster (compute).

Compute

  • You want/need to be doing responsible research.
  • Insufficient memory in desktop/laptop computer.
  • Certainty that memory errors won't invalidate your results.
  • Running research workflows on your computer means it is unusably slow for other tasks.
  • Power outages (e.g., RCD testing) are killing jobs running on your desktop/laptop computer.
  • Increasing your computational research productivity
    • Running many jobs at the same time.
    • Scaling of parallel processing enabled (e.g., MPI) software.
    • Increasing your computational research productivity through use of GPUs (from June-2018).

Platforms (web services, Rstudio, Jupyter, etc.)

  • Investigate options that are provided by others rather than re-inventing the wheel on JCU's research infrastructure.
  • JCU HPC have a very limited resource for provisioning of virtual machines for platforms that aren't available elsewhere.
  • As a service owner, you are responsible for ensuring that the services built are compliant with all JCU security/risk requirements.
  • Virtual machines can be provisioned to provide an alternative to using personal computing devices for your research (including high-end graphical work).
  • Many of the storage/compute benefits mentioned above are applicable for virtual machines.

Example Scenarios

The aim of this section is to highlight some of the values of using HPRC services can provide to a JCU researcher.  The content contained within this page and the links below is mostly a high-level view of HPRC.

Scenario 1

You have years of research work stored on your personal/work computing device.  You arrive at work one morning, power on your computing device - nothing happens.  After engaging with IT Services & Support staff, you are informed that your disk has failed and that there is almost no chance of recovering your files.  This has happened to quite a few JCU staff and students - it could happen to you.  So, you might ask "what can I do?".

HPRC - Protect your research data

Scenario 2

You have amassed or inherited a vast collection of research data in your time.  One day it dawns on you how scattered your data is - work and personal computing devices, USB disks, cloud storage providers (e.g., DropBox), etc.  The reason this has come to be is that your research data has slowly grown over time and you've purchased extra storage only when needed.

Be aware that I have received USB disks from researchers that had unreadable sectors.  HPC staff do not have any special equipment/software to try to extract data affected by faults on disk hardware.

HPRC - A location for all your research data

Scenario 3

You don't feel that investing time learning to use HPC is worthwhile - running jobs sequentially on your desktop computer is good enough.  You've recently scaled up your model and your computing jobs are taking days (or longer) to finish.  You are frustrated by power faults (black-outs, brown-outs, or surges) killing your jobs, meaning you have to start over.  Trying to remember which part of your data you were working on is hard enough, without trying to remember any special options you provided to the software you were using.

Another situation that I know has frustrated some researchers in the past, has occurred when using a laptop to run your jobs.  Your starting to tire and really keen to get home and relax, but you have to wait for a job to finish without know how long that might be.

As a general rule, HPC infra

Scenario 4

A deadline is looming and you are almost certain that running jobs on your personal computing device(s) will not get the results out in time, without significantly affecting other things you use your device(s) for.   You have overheard conversations that other researchers are using HPC and running lots of jobs at the same time.  If only you had looked at HPRC as a viable option earlier, you wouldn't be stressing about missing the deadline that is looming over your head.

HPRC - Improve efficiency by parallelising your research computing jobs

Scenario 5

You are looking to share and collaborate in modern ways.  You have seen products that suit your needs.  Technology solutions staff often provide guidance/assistance with requirements gathering and may even be able to build platforms that fulfil the requirements, subject to availability of resources and costs.

  • No labels