The HPC components of Mahuika (Cray CS400) and Maui (Cray XC50) have been designed to handle fundamentally different workloads:
- Small (one to a few) core count jobs;
- Small node count jobs (up to 26 nodes (936 cores) in an InfiniBand island);
- Embarrassingly parallel jobs at either the core level or at small node counts;
- Jobs that have poor I/O behaviour (e.g. require high (filesystem) metadata services);
- Jobs needing access to huge memory resources on one node.
- Multi-node jobs that demonstrate high performance I/O behaviour (e.g. primarily streaming I/O);
- Jobs that have good scaling properties;
- Embarrassingly parallel multi-node jobs.
When we receive your application, it will be considered by a technical expert who will make internal recommendations as to:
- Whether your workflow is suitable for any NeSI computing facility,
- If so, which facility (Pan, Māui or Mahuika) would be most appropriate,
- An estimate of how many CPU core hours (and over what period of time) your work will need.
These recommendations will inform our subsequent decision making on your application.
When considering which facility is most appropriate, in addition to the system requirements and behaviour of the software, we may have regard to:
- Other commitments of HPC resources we have made, or are likely to make during the period of your allocation;
- The entitlements of your institution to particular HPC facilities.
During the technical review process, our assessor may contact you for more information regarding your workflow, software, data storage, data transfer, memory and other requirements. Your prompt feedback will help us to complete a timely technical assessment.