2.3 Exascale—How Are We Going to Get There from Here?

Tuesday, 9 January 2018: 11:00 AM
Room 10C (ACC) (Austin, Texas)
Marc Cotnoir, CSRA, Inc., Fairfax, VA; and R. Hood

Exascale is widely considered to be the next paradigm shift in high performance computing. Like previous paradigm shifts (for example, the transition from serial to vector in the ‘80s and from vector to parallel in the ‘90s), the transition to exascale systems will present challenges for applications. We as a community and as individual developers or development teams will have to meet these challenges if we are to gain the benefits of exascale – and if we are to meet our goals for the science we want to do with next generation weather, water and climate models.

This presentation will present a perspective on what exascale systems may look like and where exascale systems will come from and then present a perspective on the expected characteristics of next generation models. Based on these perspectives, we will identify how exascale will affect applications and the challenges exascale will present for application developers. Then we will outline what it will take for the weather, water and climate modeling community to bring future models to exascale systems so as to gain the promised increases in performance.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner