Fujitsu Receives Order for World's Fastest Linux Cluster System from Japan's Institute of Physical and Chemical Research

Fujitsu Limited today announced that the company has received an order from Japan's Institute of Physical and Chemical Research, known as Riken, for a supercomputing-grade cluster of Linux™ computers with a total of 2,048 CPUs capable of achieving 12.4 teraflops⁽¹⁾ of computing performance. This system will be a model for the next generation of large-scale computing centers, based on Linux, grid computing, and other advanced technologies. It is scheduled to be operational by March 2004.

The new system will be used mainly for bioinformatics applications. In recent years the field of biotechnology has seen an explosion in the volumes of data to process. Compared to the most advanced systems currently used for homology searches⁽²⁾ of genome sequences, or for analyzing the structures and functions of proteins, the new system is expected to achieve a seven-fold performance increase.

With the growing popularity of Linux and the increasing power of personal computers, PC-based cluster systems have become an attractive option for high-performance computing, especially in academic and research institutions, as performance is scalable by simply increasing the number of processors. Cluster systems make up a quarter of the 500 fastest computing systems in the world today.

But because of the technological sophistication required for employing cluster systems, to date the application of such systems has been limited to individual departments or projects.

The system ordered by Riken will make use of grid computing and other advanced technologies, and will form the basis of a next-generation large-scale computing center that can be used by anyone at Riken's dispersed research facilities such as its central lab, or labs in Yokohama, Kobe, and Harima. Furthermore, the system will be linked as a computing node to the Information Technology Based Laboratory (ITBL)⁽³⁾ being sponsored by the Ministry of Education, Culture, Sports, Science and Technology (MEXT), and it is expected that all institutions participating in ITBL will be able to share computing resources. The system is expected to serve as a model that will spur private-sector uptake of PC cluster systems.

In realizing this high-availability, high performance parallel computing environment, Fujitsu incorporated a number of technologies such as compiler and job control functions, which were developed for its VPP⁽⁴⁾ systems; high-speed InfiniBand™ ⁽⁵⁾ interconnect in addition to grid computing applications.