80 Years of Computing: From Konrad Zuse to Exascale Computing
Thursday, June 3, 2010, 11:00am – 1:00pm, Hall B
- Prof. Dr. Wolfgang Karl, Professor, Karlsruhe Institute of Technology, Germany
- Dr. Andy White, Deputy Associate Director for Theory, Simulation & Computation, Los Alamos National Laboratory, USA
In this unique session we are reflecting on 70 years of computer history and will take a peak at the next 10 years of developments in high performance computing. Those seventy years span a performance from the Flops to the Petaflops. In the nest 10 years we expect another increase in performance by a factor of 1000 and reach the Exaflops level.
At the beginning of the extraordinary journey of growth in computing power stands Konrad Zuse, one of the pioneers of computing. He would have celebrated his 100th birthday in 2010. The ISC conference will honor Prof. Zuse with a special lecture presented by his son, Prof. Horst Zuse.
11:00am – 11:45am
Thursday Keynote Talk: TOP5 List of the Early Computers
Prof. Dr.-Ing. Horst Zuse, Professor, TU Berlin/FH Lausitz, Germany
11:45am – 1:00pm
Exascale Computing Panel
- Prof. Jack Dongarra, University of Tennessee and Oak Ridge National Laboratory, USA
- Prof. Thomas Lippert, Research Center Jülich, Germany
- Prof. Satoshi Matsuoka, Tokyo Institute of Technology, Japan
- Prof. Dr. Thomas Schulthess, Swiss National Supercomputing Center, Switzerland
- Prof. Thomas Sterling, Louisiana State University, USA
After entering the era of Petaflops computing in 2008, when Roadrunner became the first system the reach the Linpack Petaflops performance, the community started to set its sights on Exaflops computing, a factor of 1000 in increase of performance over Petaflops. Simple straight-line extrapolations of the #1 system data in the TOP500 indicate that the first Exaflops system could become available in 2019. But an initial discussion of the topic such as the Exaflops town hall workshops in the US, the DARPA Exascale study, and the extreme computing applications workshops indicate that it will be much more difficult to reach Exaflops computing. There are major challenges ahead in terms of increased parallelism and high power consumption. In this panel several experts, who have worked on the initial formulation of Exascale computing programs will share their opinions about the obstacles ahead on the road to Exaflops.
Questions for the panelists:
- There seems to be consensus in the US that an energy efficient Exascale computer by 2019 can only be built with a major investment into new technologies that could require the level of several billion dollars. What application of Exascale computing could justify such a huge investment?
- A straight line LINPACK projection of the TOP500 data indicates an Exaflops system to be become available by 2019. Do you think that is a realistic projection? What are in your opinion the biggest challenges that could slow down the development of Exascale systems? Do you think it will be possible at all to build an Exaflops system in 2019 that will require (say) "only" 20 MW?
- What new technologies do you think need to be developed to accomplish the 20 MW Exascale system by 2019?
- Will the applications be ready to run on an Exascale system? What needs to be done in the next decade to develop Exascale applications?
- What are the challenges for the software environment of an Exascale system? How do you expect an Exascale system to be programmed?