Fujitsu Deploys Japan’s Fastest Clustered HPC System

Fujitsu on Aug. 7 revealed that it has completed deployment of a new clustered supercomputer for Japan’s Institute of Physical and Chemical Research, or RIKEN, that the company claimed was benchmarked as the fastest HPC system in Japan.
Fujitsu said the new supercomputer system achieved a LINPACK benchmark performance of 97.94 teraflops. Compared with other cluster systems included in the June world Top500 list, the system’s performance ranks first in Japan and 34th in the world, the company said.
The IBM BladeCenter-clustered Roadrunner HPC system, located at the Department of Energy’s Los Alamos National Laboratory, in Los Alamos, N.M., is currently benchmarked as the world’s fastest computer system.
The BladeCenter cluster can process up to 1.105 petaflops (quadrillions of floating-point operations per second). The system in June 2008 became the first to break the petaflop LINPACK barrier.
The LINPACK benchmark is the universal standard used by the Top500 project, which compares the performance of supercomputers and publishes results every six months.
The new Fujitsu system, called the RICC (RIKEN Integrated Cluster of Clusters), comprises three computing subsystems: a massively parallel PC cluster, a large-capacity memory server and multipurpose parallel PC cluster. Each computing subsystem is connected to a common front-end system, and disk and tape systems.
The massively parallel PC cluster is made up of 1,024 Fujitsu Primergy RX200 S5 PC servers, with 2,048 CPUs (8,192 cores), Fujitsu said.
Other components in the operating environment include Parallelnavi, Fujitsu’s HPC middleware that integrates the system management functions, plus a high-speed parallel file system and program-development environment.