High Performance Computing

High Performance Computing (HPC) uses specialist systems with large amounts of cores (CPU/GPU), lots of memory, fast interconnects and fast storage. They are mainly designed for intensive parallel processing workloads.

The Research Technology Services group provides access to, and support for, a range of HPC systems including the UNSW shares of National Computational Infrastructure (NCI) and local shared cluster Katana.

UNSW Shares at NCI

 
We are providing  UNSW researchers for 2018 with about 100m SU (compute core hours). Different schemes are available:
 
NCMAS 
 
This scheme is run by NCI and offers access to Raijin@NCI, Magnus@Pawsey, MASSIVE and FlashLite. The application period is from September to October. Results are announced in December. Resources can be accessed the following calendar year. Details are here.
 
***Slides from Roger Edberg's talk explaining how the NCMAS application round for 2019 works can be downloaded from here.***
 
This scheme is run by RTS and offers access to Raijin@NCI. The application period is from November to December. Results are announced in December. Resources can be accessed the following calendar year. Smaller applications (less than 20k SU/quarter) can be made at any time.
 
UNSW LIEF
This scheme is run by RTS and offers access to Raijin@NCI to major HPC groups at UNSW.
 
Intersect 
 
This scheme is run by Intersect and offers access to Raijin@NCI. The application period is from November to December. Results are announced in December. Resources can be accessed the following calendar year. Smaller applications (less than 20k SU/quarter) can be made at any time. Details are here.
 
Note
The UNSW, UNSW LIEF are subject to mid-term adjustments. We review all project in the middle of each quarter. If a project has not used at least 30% of it’s allocation it is subject to adjustment. An email to the CI will be sent to inform about the pending adjustment.
 

How to apply

This is straight forward and identical for UNSW, UNSW LIEF and Intersect. Log on to Mancini on the NCI web site.
 

UNSW Systems

 

Katana is our shared system and provides 35 Million SU per year. It is free to use by UNSW researchers and also has a buy-in model where additional compute “blades” can be purchased by research groups.

Frequently Asked Questions

We are providing  UNSW researchers for 2018 with about 100m SU (compute core hours) on Raijin@NCI and to a small part on Magnus@Pawsey. Different schemes are available. Which should you use?

 
  1. Big HPC users (>400 kSU/year) are encouraged to apply under NCMAS and supplement with one or more of the other schemes.
  2. Normal HPC users should normally use the UNSW scheme.
  3. Users who apply outside the normal application period should apply under the UNSW or Intersect scheme.
  4. Users who need access to other machines than Raijin (Magnus or Massive) should apply via NCMAS.

 

Are there attribution policies that I should be aware of when I use any of these systems?

 

  1. Intersect - The attribution policy is available at http://intersect.org.au/attribution-policy.
  2. NCI - The attribution policy is available at http://nci.org.au/about-nci/our-role/how-to-acknowledge-nci
  3. UNSW Scheme - Any papers should have the following citation added "I acknowledge Research Technology Services at UNSW Sydney for supporting this project with [compute, storage, consulting] resources".
  4. Katana -  Any papers should have the following citation added "This paper includes results produced on the computational cluster Katana at UNSW Sydney".
1