NSCI Seminar: Ten Years of Exascale – Are We There Yet?

The first set of exascale workshops were held ten years ago in the Spring of 2007. Very quickly it was recognized that reaching the next big performance level in high performance computing is not just a matter of scaling, but that fundamental breakthroughs in architecture, software, algorithms, and application were required.  But at the time and very optimistically 2018 seemed to be the year exascale was achievable. In the years from 2008 to 2010 more workshops refined the path towards exascale computing in the U.S., and by about 2010 delivered the outline of a plan on how to achieve this goal. Also internationally several countries developed exascale initiatives. So all indicators were positive. Yet now, ten years later in the second half of 2017,  we are still about five years away from exascale and 2023 seems to be a more likely target. What happened? I will revisit the original goals and projections, and analyze where and when the community started to fall behind. This will be a useful analysis to draw lessons for the next five years, and see how exascale computing can eventually be achieved.