Despite the recent headlines that Ireland now ranks highest in number of Top500 supercomputers per capita, we have continued our steep fall on the research High Performance Computing stage – a trajectory that has been highlighted on several occasions. It is now 3.5 years since Ireland had a research facility ranking on the list – the Irish Centre for High-End Computing (ICHEC) Fionn installation. While Ireland does claim 4 Top500 machines on the current Top500 list these machines are owned and operated by private industry enterprises and do very little if anything to promote research in Ireland. As demonstrated by the log-scale graphs below of the largest research/academic machines in several European countries, Ireland compares extremely unfavourably against other European countries. On a global stage, where two counties alone (China and the US) account for just over 2/3 of installations and very nearly 2/3 of performance the picture is even more lopsided.
The largest research installation in Europe is currently the “Piz Daint” machine with an impressive 361,760 cores and a performance of 19,590.0 TFlops/s. Ireland’s largest machine Fionn, managed by ICHEC, has by comparison 7,680 cores and a performance of 140.4 TFlops/s – less than 1% of Piz Daint’s performance. Even compared to countries with more modest HPC resources than Switzerland, Ireland still fairs poorly. Poland’s largest research facility has more than 55,000 cores (1,670.1 TFlops/s) while Spain’s largest has over 150,000 cores (6,470.8 TFlops/s).
In Ireland, it is evident that ICHEC is chronically underfunded, and without increased investment Ireland will continue to occupy the bottom rungs of the HPC research ladder and the work of Irish researchers will continue to suffer. Ireland’s lack of investment in research HPC has meant that the country sits at or close to the bottom in terms of meaningful metrics that account for HPC resources openly available to researchers. For Irish researchers this leads to longer queue times and HPC facilities that are increasingly unable to support world-class research.