The research area aims at developing system software technologies as well as related systems to be used for high performance computing in the post generations of the Japanese national supercomputer K.
More concretely, research and development will be conducted for system software enabling us to exploit maximum efficiency and reliability from supercomputers which will be composed of general purpose many-core processors as well as special purpose processors (so called GPGPU) in the second half of (and/or after) 2010's. In addition to the system software such as programming languages, compilers, runtime systems, operation systems, communication middleware, and file systems, application development support systems and ultra-large data processing systems are the targets for research and development. Also, the targets include system software in the overlapping layers of software stack, which encourages real usages of developed system software.
Development of System Software Technologies for post-Peta Scale High Performance Computing Research Supervisor:Akinori Yonezawa (Professor, Graduate School of Information Science and Technology, The University of Tokyo)
Numerical simulation and data analysis utilizing super-scale computation and data processing are now regarded as the third methodology of science, which will play critical roles after the first and second methodologies, namely, theory and experiment/observation. Accordingly, Europe, US and China are engaged in severe competition in developing most advanced supercomputers. Japan has also launched the national project of developing the next-generation supercomputer called the K supercomputer and it will be fully installed in 2012. Reflecting the importance of supercomputers, it has been widely recognized that these countries, behind the scene, have already begun their projects of developing next-next generation supercomputers.
Assuming application software for supercomputers in various science and engineering disciplines, it is indispensable to be equipped with well-designed, highly functional and reliable system software which allows us to exploit the maximum performance of supercomputer hardware. In our research area, we will conduct research and development for system software such as programming languages, compilers, runtime systems, operating systems, communication middleware, and file systems as well as application development support systems (including numerical libraries) and ultra-large data processing software systems.
So far the future direction of supercomputer architectures has not been clear except that many-core general purpose processors and/or special purpose processors (so called GPGPU) will be components of the next-next generations supercomputers. Thus, our research area requires applicants to submit research proposals which state the assumed supercomputer architectures as clearly as possible. In the proposals, it is also required to show that system software to be developed will probably work efficiently on the assumed architecture. System software to be developed needs to be made public as open source software and contribute itself to the progress of the research area. Furthermore, it is desirable that system software to be developed enables targeted applications to run efficiently in accordance with the underlying architecture as well as lower system software layers.
Assuming the maximum research period being 5 years long, at the interim review process, we will evaluate research in terms of possibility of real use of the system software resulting from the research. In case the evaluation result of a research project does not meet criteria, its continuing research plan will be substantially revised, it may be requested to be merged with other projects, or it may be terminated.
By promoting the research area, it is expected in 2015 or after that the basis of system software technologies will be created to be used for post-peta scale supercomputer systems after the Japanese next-generation K supercomputer. Also it will contribute to directing the development of system architectures as well as software architectures for the next-next generation massive parallel supercomputing.
For these purposes, we encourage research projects to share information with researchers in international and industrial sectors, and form research organization structures keeping industry-academia and international relations close.
At the end of research period (around 2016), it is expected for the research area to produce the system software to be utilized on massively parallel computing systems, and furthermore it is expected to bring new developments of science and technology with innovative breeds of simulation and prediction methods of ultra-large scales.
From this fiscal year, we would like to expect researchers in the fields of HPC applications, embedded computing, computer architecture and alike to apply for our research area and conduct research of system for software. Further we would like to have proposals from young researchers who would be the core in the era of Post-Petascale. Also we welcome proposals of smaller scale, 30 to 60 million yen per year (Category 1) which still meet 5 year research plans of research directors.