Tutorial II: High-end computing systems: Past, present and future
Hiroaki Kobayashi · Society of Instrument and Control Engineers of Japan · 2012
High-end computing systems so called Supercomputers are now an important R&D infrastructure to advance science and engineering such as Earthquake and Ts u n a m i analysis, Climate Prediction, Plasma Physics, Computational Biology, automobile design and etc. The first supercomputer was developed in 1975, and since then, the performance of supercomputers improves at a rate of 10 times ever four years (or 1,000 times every 10 years). In this talk, the technology and architectural innovations that contribute to this skyrocketing improvement of supercomputers will be given by introducing modern supercomputer systems such as vector, scalar and GPU clusters. In addition, the on-going national R&D project, named “innovative HPCI (High-Performance Computing Infrastructure),” will be briefly described. HPCI is a nation-wide supercomputing infrastructure that makes K-computer, which is ranked No.1 in the current TOP500 supercomputer ranking, and other national supercomputers accessible in an integrated manner (single-system image), and will be developed and deployed this year. Finally, the outlook for exascale-computing systems expected to be available by the end of this decade will also be discussed.