Getting the most out of your NeSI allocation
At New Zealand eScience Infrastructure (NeSI), we want to help you run your work as efficiently as possible on our platforms. The benefits of this are two-fold: on one hand, it could reduce your time to solution, thereby improving your research productivity, and on the other hand, improving efficiency could reduce queue times for all users of the platform.
In this talk, we will show how you can identify whether your jobs are fully utilising the hardware they have been allocated and also give some examples showing how we have helped some researchers improve the efficiency of their jobs on NeSI. This includes investigating different compilers and compiler flags, run time tuning including exploring different options for parallelism and investigating the scalability/parallel efficiency of your jobs. In some cases, we have achieved significant improvements by making relatively small changes.
ABOUT THE AUTHORS
Chris Scott is a Research Software Engineer for NeSI based at the University of Auckland.
Anthony Shaw provides Applications Support for NeSI based at the University of Otago.