AIST Super Cluster

Outline of Project

The AIST Super Cluster consists of more than 3000 processors and is the largest-scale cluster system in Japan. It is being introduced as part of the establishment of the Industrial, Academic and Governmental Cooperative Information Technology Research Facilities for creating new industries and venture businesses. It started operation in March 2004.

The AIST Super Cluster consists of many computers, so-called computing nodes, connected through high-speed networks. It makes use of grid technology for flexible access of information on the networks and forms the basis of high-performance computing environments. It is intended to establish the technology of constructing large-scale cluster systems with thousands of connected processors and to be used in nano-technology, bioinformatics and other research fields. It serves as the core system for inter-field and international research promotion and cooperation between industries, academy and government.

This system consists of three cluster systems and a 20-TB storage system. The three cluster systems are the "P-32 Cluster System" featuring total computing performances, the "M-64 Cluster System" featuring a large memory cluster for each computing node, and the "F-32 Cluster System" designed for simultaneous execution of several independent computations. In particular, the P-32 Cluster System and M-64 Cluster System use the Myrinet networks exclusively for computation. These three cluster systems are connected through ultra-high-speed 10-Gbps networks. The cluster system as a whole flexibly covers various types of engineering computations. It is especially suitable for the following:

Computation that requires high processing capability using a large number of high-speed processors such as computations in the bioinformatics field, although not requiring a large storage capacity of each computing node comprising the cluster
Computation that requires a large storage capacity of each computing node and high processing capability of the processors such as computations in the nano-technology field
Computation that requires independent processing but does not require mutual communication, as in parameter surveys with many case studies, for example shape deformation of buildings and vehicles or addition of parts.

The AIST Super Cluster is expected to serve as a mighty tool in research fields requiring large-scale computing capabilities.

Terminology

Cluster system: One of the distributed memory computer systems with several computers connected through a high-speed network. It frequently employs relatively low-cost computation servers with commodity processors as the component units (computing nodes). Each computing node has several processors and a main memory and runs on an independent OS. The computing nodes execute large-scale computations in cooperation with each other. The cluster system is used for scientific/technological computations that cannot be performed by a single computing node and load distribution processing such as parallel databases.