SS-ken HPC Forum 2009
System and application for Petascale Computing
- 1.Date ： 2009/9/3(Wed)
-
・Forum 10:30〜18:20
・Reception 18:40〜20:00 - 2.Place ： Shiodome City Center, 24th floor; Fujitsu Ltd.
- 3.Overview
-
Recently, with the increasing needs for large-scale computing in the various science and engineering areas like nano-science and materials, life science, meteorology and disaster prevention, environment and energy, nuclear power, and aerospace as well as with the progress of the practical application of advanced simulation and analysis in industries aiming at reducing cost and leading time for the product development, we are entering into a new era in which a peta-flop-level computing performance is substantially required. However, for accelerating the practical use of peta-scale computers, breakthroughs more than what it had been and further scales-up are required not only in the field of the processor and system technology but also in the field of parallel programming, tuning, file systems, middleware and technologies in large-scale pre and post data processing.
Given this situation, as the theme "Systems and applications in the peta-scale era", invited lectures on topics of hot trends and challenges about peta-systems and applications will be given by experts. In the International Invited Talk at the beginning, Prof. Jack Dongarra from the University of Tennessee, who is famous for Linpack and TOP500, gives a talk about HPC in terms of the current trends and challenges for the future. Finally, FUJITSU holds an introduction of their efforts for the period of peta scale.
Through the lectures and related discussion, we would like to investigate deeply on how the peta-scale applications should be, and what the problem solving and the future direction on the issues such as the development environment and large-scale parallel programming technology should be in view of ongoing peta-level project and technological innovation. - 4.Program
-
10:00 Doors Open 10:30-10:35 Openning
Murakami Kazuaki (Kyushu University)10:35-11:40 [1] Current Trends in High Performance Computing and Challenges for the Future
Jack Dongarra（University of Tennessee and Oak Ridge National Laboratory）[Abstract] In this talk we examine how high performance computing has changed over the last 10-year and look toward the future in terms of trends. These changes have had and will continue to have a major impact on our numerical scientific software. A new generation of software libraries and algorithms are needed for the effective and reliable use of (wide area) dynamic, distributed and parallel environments. Some of the software and algorithm challenges have already been encountered, such as management of communication and memory hierarchies through a combination of compile--time and run--time techniques, but the increased scale of computation, depth of memory hierarchies, range of latencies, and increased run--time environment variability will make these problems much harder. We will focus on the redesign of software to fit multicore architectures.
[Keyword] High Performance Computing, Multicore, numerical software, algorithms, parallel computing
11:40-12:40 [2] The Multiscale Multiphysics Heart Simulator Aiming at Practical Application
Hisada Toshiaki[Abstract] (Under Preparation) A heart simulator has been developed at the University of Tokyo based on the transdisciplinary research on computational science (Hisada) and medical science (Sugiura) since 2001. During 2003-2008 'Development of Multiscale Multiphysics Heart Simulator for Medical Science and Drug Discovery' was conducted under the JST program, CREST. Since 2007 'Application of Heart Simulator to Medical Treatment' has been driven in collaboration with Application Research and Development Division, Next Generation Technical Computing Unit, Fjitsu Ltd (Division Manager Kadooka) under the JST project, Collaborative Development of Innovative Seeds-Practicability Verification Stage. The state-of-the-art of our heart simulation and its significance will be addressed.
[Keyword] Heart Simulation, Multiscale-Multiphysics, Finite Element Method, Fluid-Structure Interaction Analysis, Parallel Computation
12:40-14:00 Lunch 14:00-15:00 [3] Overview of JAXA Supercomputer System (JSS) and its performance assessment
Takaki Ryoji（Japan Aerospace Exploration Agency）[Abstract] Japan Aerospace Exploration Agency (JAXA) has long recognized the importance of numerical simulations using high performance computers and has strongly promoted the installation and operation of large scale high performance computing systems. In April 2009, a new system, called JAXA Supercomputer System (JSS) was installed and started operations. This system consists of several computer systems and the main system is Fujitsu FX1, a multicore based scalable parallel cluster with approximately 12,000 cores, having 120 Tflops peak performance and 94 TByte of main memory.
This paper reports the results of preliminary performance evaluation with our current aerospace CFD applications as well as overview of the new system.[Keyword] aerospace, CFD(Computational Fluid Simulation), scalable parallel cluster, multicore
15:00-16:00 [4] Fusion plasma simulations and their massively-parallel benchmarks
Watanabe Tomohiko（National Institute for Fusion Science）[Abstract] One of the central subjects in the magnetic confinement fusion research is comprehension, evaluation, and prediction of the turbulent transport that is crucial to confinement property of high-temperature plasma. We have advanced large-scale kinetic simulations of the plasma turbulent transport, where dynamics of particle distribution functions in the five-dimensional phase-space are considered. I would briefly introduce the current status, present benchmark results of fusion plasma simulation codes on several massively-parallel supercomputers, and discuss the future directions.
[Keyword] Fusion, Plasma, Numerical Simulation, Massively-parallel Computation
16:00-16:15 Intermission 16:15-17:15 [5] Large-Scale Parallel Strategies in an Open Source FE Structural Analysis Software FrontISTR (tentative)
Okuda Hiroshi（The University of Tokyo）[Abstract] Supported by the Ministry of Education, Culture, Sports, Science and Technology (MEXT). the Center for Research on Innovative Simulation Software at the University of Tokyo has been promoting the project of "Innovative Simulation Software". FrontISTR is an open source FE structural analysis software being developed at one of the subproject of "Development of the Next Generation Manufacturing Simulation System". Products in its practical form are the assembled structures, which are composed of various parts and blocks. Nonlinear static and dynamic behavior of the assembled structures as a whole can be analyzed by FrontISTR. Based on a hierarchical mesh refinement and a hierarchical convergence algorithm, very large scale models with up to 10 billion elements will be handled.
[Keyword] Structural Analysis, Parallel Finite Element Method, Assembled Structures, Hierarchical Mesh
17:15-18:15 [6] Progress report on Fujitsu's peta-scale supercomputer
Inoue Aiichiro（Fujitsu Ltd）[Abstract] Fujitsu has announced a new processor for HPC in May. This CPU has been developed for peta-scale computers. It takes features from SPARC64(TM) series which are the heart of Fujitsu's reliable business servers. In this talk, new features of the high performance and low power consumption CPU, SPARC64(TM) VIIIfx, will be introduced. At the same time, new software technology to utilize this CPU's capability to support hybrid parallel programming will be explained. And a Fujitsu's roadmap for future Technical Computing including PC cluster will be shown.
[Keyword] high reliability, high performance, low power consumption, SPARC architecture, hybrid parallel
18:15-18:20 Closing
Matsuo Yuichi (Japan Aerospace Exploration Agency)18:20-18:40 Intermission 18:40-20:00 Reception（\500） - 5.Exhibition： FX1、PRIMERGY BX900、SPARC64 VIIIfx
- 6.Registration fee： Free （Reception: \500）
SS研facebookページへ