Tokyo, February 19, 2008
Fujitsu Limited today announced that it has received an order from the Japan Aerospace Exploration Agency (JAXA) for a supercomputer system using Fujitsu's FX1 technical computing server.
At the core of the new supercomputer is a massively parallel computer system comprised of 3,392 FX1 computing nodes, delivering peak theoretical performance of 135 teraflops (TFLOPS)(1). This system will replace the agency's existing supercomputer system, called the Numerical Simulator III.
JAXA extensively uses advanced numerical simulation technology in the development of rockets, advanced space-transport systems and aircraft. This new supercomputer system will serve as the core computing engine for JAXA and is expected to give the organization an important boost in its R&D activities.
The transition from the current system to the new one will begin in April 2008, and the entire system is to be up and running by April 2009.
Supercomputer Utilization at JAXA
Beginning in 1987, JAXA set out a computer simulation plan with the goal of developing and disseminating computer-based numerical simulation as a tool for the aerospace industry. Since 2002, JAXA has been using a supercomputer system with Fujitsu's PRIMEPOWER HPC2500 servers at its core. This system, known as the Numerical Simulator III, has peak theoretical performance of 9.6 TFLOPS and 3.6 terabytes (TB) total memory. As the organization's R&D activities have grown in scale, complexity, and diversity, its needs for computing power have grown dramatically, leading to its decision to install a new system.
As part of the JAXA Vision 2025 (2), JAXA is pursuing research to enable the development of next-generation rockets with world-class performance, reusable future space transport systems, and aircraft, as well as research to enable exploration of the Moon and other planets. In such uses as early assessment and design support in exploring new designs or in the development stage, or in simulating the environment of space, which cannot be reproduced on Earth, the new supercomputer system will improve the efficiency and reliability of the development process and help spark the creativity of researchers and developers.
Features of the New Supercomputer
At the core of the new system is a massively parallel computer system comprised of 3,392 FX1 computing nodes. Compared to the existing system, the new system delivers peak theoretical calculating performance of 135 TFLOPS, approximately a 15-fold increase, 100 TB of total memory, a roughly 30-fold increase, and total storage of 11 petabytes (PB), approximately a 16-fold increase.
In addition to that system, there will be multiple subsystems, including a vector computer and a scalar SMP computer, to enable numerical simulations that take advantage of the massive memory space, offering users an optimized and efficient computing environment that can accommodate increasingly diverse R&D activities.
The new system will enable the realization of the JAXA Vision 2025 of advancing and disseminating numerical-simulation technologies for every area of research and development.
Comment from Kozo Fujii, Director, Engineering Digital Information Center, JAXA
"The new system is the first supercomputer system JAXA has ordered since its formation through the merger of three space and aerospace research institutions. In addition to pursuing aviation research, as we have done up until now, we will apply the new system in the fields of rocket engine analysis and rocket plume acoustic analysis as well as for spacecraft design. In these applications, we aim to make significant contributions by using and developing the system for the fields of space science, development, and exploration."
About the FX1, a Next-Generation Technical Computing Server
The FX1 employs a new computing architecture to get the most from its multi-core CPU capability. To maximize that capability and the performance Fujitsu's SPARC64™ VII high-performance quad-core CPU, they are combined with a high-performance compiler and the company's own custom chipset designed for well-balanced, high memory bandwidth, enabling an extraordinarily high level of computing efficiency that is difficult to achieve with conventional multi-core systems. In addition, through intelligent interconnects(3) that enable various types of calculation to be performed on data in the transmission pathway, a significant leap in the scalability of parallel applications is achieved. Moreover, the rich suite of development platform tools and operational functions delivered by Parallelnavi, the parallel-processing software that has a long track record in the field of technical computing, makes it easy to work in a high-performance parallel computing environment.
Attachment
FX1 Key Features & Specifications (67 KB)
