Most people use their computers (which includes mobile phones) for communication, social media, games, entertainment, office applications, and the like. Most of the time these activities are not particularly onerous in terms of computing as such or do not lead to enormous benefits in productivity, inventions, and discovery. There is one field, however, rarely discussed, that does do this - and that is supercomputing. It is through supercomputing that we are witnessing the most important technological advances of our day, including astronomy, weather and climate forecasting, materials science and engineering, molecular modeling, genomics, neurology, geoscience, and finance - all with numerous success stories.

Usually, I draw a distinction between supercomputing and high-performance computing. Specifically, a supercomputer is any computer system that has exceptional computational power at a particular point in time, many (but not all) of which are measured in the bi-annual Top500 list. Once upon a time dominated by monolithic mainframes supercomputers, in a contemporary sense, are a subset of high-performance computing, which is typically arranged as a cluster of commodity-grade servers with a high-speed interconnect and message-passing software that allows the entire unit to be treated as a whole. One can even put together a "supercomputer" from Raspberry Pi systems, as the University of Southhampton illustrates.

How important is this? For many years now we've known that there is a strong association between research output and access to such systems. Macroeconomic analysis shows that for every dollar invested in supercomputing, there is a return of forty-four dollars in profits or cost-savings. Both these metrics are almost certainly going to increase in time; datasets and problem complexity are growing at a rate greater than the computational performance of personal systems. More researchers need access to supercomputers.

However, researchers do require training to use such systems. The environment, the interface, the use of schedulers on a shared system, the location of data, is all something that needs to be learned. This is a big part of my life; in the last week, I spent three days teaching researchers from the basic of using a supercomputer system to scripting jobs, to using Australia's most powerful system Gadi at NCI, along with contributions at a board meeting of the international HPC Certification Forum. It is often a challenging vocation, but I feel confident that it is making a real difference to our shared lives. For that, I am very grateful.