Thursday 24 March 2011

Investments Today for Effective Exascale Tomorrow

I contributed to this article in the March 2011 The Exascale Report by Mike Bernhardt.

"Initiatives are being launched, research centers are being established, teams are being formed, but in reality, we are barely getting started with exascale research. Opinions vary as to where we should be focusing our resources.

In this issue, The Exascale Report asks NAG's Andy Jones, Lawrence Livermore's Dona Crawford, and Growth Science International's Thomas Thurston where should we (as a global community) be placing our efforts today with exascale research and development?"

THE EXASCALE REPORT ASKS: Where should we (as a global community) be placing our efforts TODAY with exascale research and development?

Here is a short version of my response (read the article for the full responses from myself, Dona and Thomas) ...  [The full article requires subscription - the extracts here are not complete and are modified slightly to support that.]

As application developers, we only know two things about (early) exascale supercomputers at the moment: (1) billion–way parallel; (2) data movement (bandwidth and latency) will be very costly. We might make assumptions about other details or features but it would just be speculation. But, if we are to have real, useful applications running on the first exascale systems from day one, then many of the application design choices will have to be made before we know any details for sure.

The global community (especially on the applications development and user side) must focus efforts today on pragmatic rather than elegant exploitation of early exascale computers. That is, find the most economical path to achieve useful science (or other important results) as soon as possible.

Why? (1) early high profile science output will be essential to help justify the (public) funding invested to that point and release further funding for the future development plans; (2) early insights from real world applications on exascale computers will be required to properly inform the development of exascale software and system technologies to ensure future generations of exascale supercomputers are broadly useful and economically viable.

No comments: