At the end of March 2004, the National Institute of Advanced Industrial Science and Technology (AIST), an independent administrative institution, installed a cluster computer system of the highest overall computing capability in Japan, named “AIST Super Cluster”
1, and has put it into operation. The AIST Super Cluster is characterized by overall computing peak performance of 14.6 TFLOPS, 9.6 TB main memory capacity and 803TB storage, and based on a Linux operating system. It has about 1/3 of peak performance of the Earth Simulator, the world’s fastest computer.
The AIST Super Cluster consists of three cluster subsystems, “P-32” attaching much weight to overall computing capability, “M-64” stressing massive memory capacity of individual computers, and “F-32” aiming at independent parallel and multiple processing, as well as 20 TB storage subsystem. These subsystems are interconnected through 10 giga-bit Ethernet, so as to optimize the overall performance through either separate or combined operations.
The AIST Super Cluster has been installed with the following intentions:
1. Constructing a high performance computing environment based on grid technology.
A high performance grid computing environment will be built up by using the AIST Super Cluster as an infrastructure to provide computer resources for the grid.
1 Website of AIST Super Cluster (English):
http://unit.aist.go.jp/tacc/en/supercluster.html
2. Establishing construction and operating technology of large-scale cluster systems.
Cluster computers have been widely used for realizing supercomputing environment at a reasonable cost through the combination of a lot of personal computers (PCs). However, technology for constructing and operating a large-scale cluster system comprising 1,000 or so PCs has not yet been established. It will be attempted to test various operational modes and to check out operating technologies by using the AIST Super Cluster composed of cluster sectors of different constitutions.
3. Providing computer resources for R&D on nano-technology and bio-informatics
Computer resources will be provided for R&D works in frontier areas such as material sciences to analyze nanometer scale events occurring at material surfaces and interfaces of different materials and biosciences to study behaviors of proteins and other biopolymers.
4. Serving a core system of infrastructure with collaborative community
Used as one of grid computing resources, the AIST Super Cluster will serve for promoting joint works with foreign organizations, industrial revitalization, and infrastructures for international and industrial-academic-governmental collaborations.