Tweaking the High Performance Computing scheduler to encourage researcher investment

Mr John Zaitseff1, Mr Martin Thompson1

1UNSW Sydney, Sydney, Australia

Many research groups like to use at least a portion of their grant funding for purchasing computing hardware, in the expectation that such capital expenditure can be utilised far beyond the lifetime of the grant.  Unfortunately, uncontrolled spending in this way often leads to under-utilised equipment, out-of-date and unsupported hardware and software and a potential for significant lapses in cybersecurity.

Research Technology Services (ResTech) at UNSW Sydney maintains a High Performance Computing (HPC) facility, Katana, that provides an ideal, controlled location for such grant spending: researchers co-invest by purchasing the computational hardware, while ResTech provides the surrounding infrastructure including networking, hardware, software and security support.

This talk shows how ResTech has tweaked the traditional HPC scheduler to encourage such researcher investment while simultaneously allowing a high level of utilisation across all HPC compute nodes.  These tweaks have contributed to a significant reduction in uncontrolled purchases of computing equipment among researchers at UNSW Sydney.


Biography:

John Zaitseff is a Research Computing Support Engineer at UNSW Sydney.  He provides both wide and deep expertise in all aspects of High Performance and Research Computing to researchers, including analysing computing needs and requirements, providing advice on purchasing and utilisation, training staff and students to use new and existing facilities, system design and administration, and performing software design, programming, debugging and code optimisation using multiple programming languages.

https://orcid.org/0000-0003-4504-1124

Categories