At DUG, we offer high-performance computing as a service (HPCaaS) for a wide variety of clients. Built on many years of experience in geophysics, our world-class compute cluster is tailor-made for scientific computing and attracts a diverse range of computationally intensive projects.
We are looking for a HPC Expert with experience in on-boarding clients, helping them to build and tune complex packages for performance on DUG’s worldwide HPC infrastructure.
As a HPC Expert, you will be part of a team of technical experts responsible for ensuring that customers can use our compute infrastructure as productively as possible. Your pragmatic approach will help you deal with third-party software that's not always especially easy to build, compile, understand and tune. You will become proficient in the languages and techniques used in these projects, as well as an evolving and somewhat-unique HPC environment.
List of Responsibilities (not necessarily exhaustive):
Benchmark customer codes against different hardware within DUG HPC Cloud
Install, port, optimise and tune codes for optimal performance on DUG HPC Cloud
Work with vendors to benchmark new hardware
Understand existing and new hardware and how it might benefit DUG HPC Cloud
SKILLS
Prerequisites:
Ability to install complex UNIX packages and dependencies
Programming ability in at least one of C, C++, Fortran
Ability to program with at least one of OpenMP, MPI, OpenACC, Cuda or pthreads
Code optimisation techniques
Strong problem solving skills
An eye for details
Familiarity with batch queueing systems
Shell scripting (eg. sh / bash / csh / tcsh)
An understanding of modern computer, network and IO architecture
An understanding of parallel IO and various parallel file systems
Experience with team collaboration and version control, ideally git
Desirable:
A true Shell scripting expert
Python scripting
Network performance and tuning
I/O performance and tuning
Memory bandwidth / cache performance and tuning
Linux kernel tuning
Familiarity with GPU programming
Familiarity with vector programming / SIMD instruction sets
Scientific programming background to understand a wide range of numerical algorithms and techniques
Experience with R programming
Experience with using SLURM
Experience teaching or delivering workshops to varied audiences
Familiarity using different compilers (Intel, GNU especially)
Familiarity with containers (especially Singularity / Apptainer)
Familiarity with metascheduler packages such as NextFlow and/or Cromwell
DISCLAIMER
The offer is subjected to pre-employment screenings that may include, but are not limited to:
Verification of your right to work in the respective location
Provision of applicable and relevant qualifications
Nationally approved criminal history check