FUJITSU Supercomputer PRIMEHPC FX10
FUJITSU Supercomputer PRIMEHPC FX10

Fujitsu Receives Supercomputer Order from Nagoya University – May 15, 2013– Tokyo, Japan  (Techreleased) – Fujitsu today announced that it has received an order from Nagoya University’s Information Technology Center for a “high-performance supercomputer for academic research.”

FUJITSU Supercomputer PRIMEHPC FX10
FUJITSU Supercomputer PRIMEHPC FX10

The system will have a hybrid configuration, composed of a FUJITSU Supercomputer PRIMEHPC FX10 and an HPC cluster comprised of FUJITSU Server PRIMERGY CX400. At deployment, it will have a theoretical peak performance of 561.4 teraflops, and will be scaled up in the future to 3,662.5 teraflops, making it one of the biggest systems in Japan and the largest in the Tokai region where Nagoya is situated.

The new system is due to start running from October 2013 and will be used for advanced research and academic purposes at Nagoya University’s Information Technology Center.

Background

Nagoya University, the largest national university in the Tokai region and a center of academics and research there, is home to the Information Technology Center, a shared resource for universities and researchers conducting academic research throughout Japan. Since December 1981, numerous researchers have used the mainframe computers and supercomputers deployed there, mostly for work on science and technology.

The new system consolidates the Information Technology Center’s three existing systems: the supercomputer system, application server, and information-academics platform. It was designed to meet demands for more computing capacity, to make computing resources in other academic areas, to create new computational services, and to help educate people who will reach into new areas of inquiry.

Fujitsu’s concept for the system was that of a hybrid, specifically, the combination of a PRIMEHPC FX10 and a cluster-type supercomputer comprised of PRIMERGY CX400 servers. The PRIMEHPC FX10 system features high computational capacity, energy efficiency, computing performance, and high availability. The PRIMEHPC FX10 should act as a bridge to exa-scale computing, performing as a computing resource under the auspices of the Ministry of Education, Culture, Sports, Science, and Technology’s High Performance Computing Infrastructure(1), thanks to compatibility with the K computer. As for the HPC cluster with Intel CPUs, it can run a wide range of ISV applications and free-software projects.

About the New System

In its initial configuration, the new system will comprise 384 PRIMEHPC FX10 nodes and 552 PRIMERGY CX400 nodes, which have a theoretical performance of 561.4 teraflops. In the future, it will be scaled up to 3,662.5 teraflops, making it one of the most powerful supercomputers in Japan and the most powerful in the Tokai region.

The PRIMEHPC FX10 supercomputer is an upgraded version of the technology used in the K computer, with outstanding scalability, reliability, and power consumption to performance ratio.

The PRIMERGYCX400 server is an x86 server suitable for HPC applications that allows high-density mounting, with 84 nodes per rack, roughly double a typical 1U server. Of the 552 nodes that this system uses, 184 are equipped with Intel® Xeon Phi™ Coprocessors, for high performance at low cost and low power consumption. This system also employs ScaleMP’s vSMP Foundation software which enables the aggregation of multiple nodes as a virtual single large compute node.

As part of HPC middleware, the system is being rolled out together with the Technical Computing Suite for peta-scale system support(2). The system has a total of 6.0 petabytes of storage, configured using the high-performance distributed FEFS file system(3) for capacity and performance.

Comment from Prof. Yoshito Itoh, Director of the Information Technology Center at Nagoya University

“As a shared resource available for use nationwide, a home for joint research, and a resource-providing institution in the HPCI, Nagoya University’s Information Technology Center chose to deploy the PRIMEHPC FX10, with high computing performance and energy efficiency, and the PRIMERGY CX400, with multi-core processors, to accelerate the use of high-performance computing in more and more fields, from initial-stage research to large-scale studies and industrial applications. It can be expected to expand and invigorate the computational-science community, which includes the next generation of researchers, and return benefits to society through scientific results.

“The new system will also comprehensively support large-scale simulations using massive storage systems and giant 3D screen virtual-reality systems. Additionally, planned system upgrades will bring sought after computational services along with even better energy efficiency.

Glossary and Notes

HPCI:
The High-performance Computing Initiative ties together major supercomputers throughout Japan, including the K computer, through a network, resulting in a computing-environment platform that can meet a wide range of needs.
Technical Computing Suite:
Fujitsu’s proprietary HPC middleware, which includes system management and job operation management functions, FEFS, as well as a compiler and library, resulting in high execution performance for massively parallel applications.
FEFS:
Fujitsu Exabyte File System. A high-performance distributed file system that enables sharing across 100,000 nodes.