SYSTEMS

Most powerful supercomputer in consortium of seven national supercomputer centers: Fujitsu Limited today announced that it has received an order for a supercomputer system from the Computing and Communications Center at Kyushu University in Japan. The center is one of seven national supercomputer centers in a consortium of universities. It offers advanced computing services for computational science, including fluid analysis and molecular science, to on and off-campus researchers. The new system will be capable of 31.5 teraflops, making it the most powerful supercomputer among the seven centers, and help meet the growing demand for scientific and technical computing at academic and research institutions. The new supercomputing system will be a hybrid consisting of two cluster systems. A large symmetric multiprocessing (SMP) cluster of 32 Fujitsu PRIMEQUEST 580 mission-critical IA servers, along with a cluster of 384 Fujitsu PRIMERGY RX200 S3 industry standard servers, will be complemented by a single PRIMEQUEST 580 acting as a file-management server. The system will run Fujitsu Parallelnavi HPC software, enabling easy operation without users having to be aware of the underlying complex hardware mix. Moreover, the system will feature the Fujitsu HPC solution. This improves usability by providing a unified monitoring environment for everything from job execution to simulations, and a web interface for managing and supervising the hybrid computing system. The new system is scheduled to become operational in June of 2007. Background of System Upgrade As a national supercomputer center, Kyushu University's Computing and Communications Center serves its own university and provides computing services to other academic and research institutions nationwide. Responding to rising demand for scientific and technical computing services the center decided to construct a 31.5 teraflops system, roughly 50 times more powerful than their existing one. This will establish the Kyushu facility as the most powerful supercomputer site among the seven centers. Overview of New System * Computation servers: 32 PRIMEQUEST 580 servers form an SMP cluster; 384 PRIMERGY RX200 S3 servers form a PC cluster (Both clusters run on Red Hat Enterprise Linux) Total combined theoretical peak performance of the computation servers is 31.5 teraflops, making it the most powerful supercomputer among the seven centers in Japan. * File-system management server: A PRIMEQUEST 580 server Running Parallelnavi SRFS for Linux, a high-speed distributed file system, the PRIMEQUEST 580 file-management server will enable the high-speed transfer of large volumes of data for the computational processing taking place, in parallel, on the PRIMEQUEST 580 and PRIMERGY RX200 S3 computation servers. Together with very reliable storage system file management it will provide the highly reliable, high-performance file system ideally suited to demanding scientific and technical computing calculation. * Storage: An ETERNUS 8000 Model 2100 enterprise disk array This enterprise class disk array storage system provides the high access speeds needed to save the computational results produced by the supercomputer computation servers. * HPC Solutions: -- Operations Management Portal: supercomputer operational management system -- This system uses a web interface to supervise and manage hybrid systems and minimize system administration tasks. Seamless visualization system -- This solution simplifies the complexity of tracking the flow from job execution to visualization. It makes supervision of supercomputer's mass data output over the network quick and straightforward. Its simplicity means even a beginner at a PC and can visualize both the job-execution results and memory used by the “in-process” jobs. This “seamless visualization” solution was developed jointly by Fujitsu and KGT Inc., a fourteen year Fujitsu partner and specialist in developing visualization software for supercomputers.