By Matthew Dublin
Japan's RIKEN Institute has rolled out a new supercomputer called the RIKEN Cluster of Clusters (RICC) thanks in large part to Fujitsu, which deployed the new machine. The RICC is capable of 97.94 teraflops performance -- making it the country's most powerful to date and 34th worldwide, according to the Top500 list. (And no, we're not forgetting about RIKEN's MDGRAPE-3, the first computer to break the petaflop barrier back in 2006, but unfortunately it is not included in the Top500 due to its specialized architecture which cannot run the requisite benchmarking software.) The High-Performance Molecular Simulation Team uses MDGRAPE-3 for intensive large-scale molecular dynamics simulations, including simulations of RNA Polymerase II and protein-protein interactions.
The folks at RIKEN are hoping that their bundle of Fujitsu joy will further crank up the productivity of a number of ongoing projects. According to Ryutaro Himeno, director of RIKEN's Advanced Center for Computing and Communication, in addition to providing RIKEN researchers with the ability "to process high-volume experimental data using advanced DNA sequencers, accelerators, and X-FEL (X-ray Free Electron Laser) ... and the ability to use the system for [advanced] computational software development using MDGRAPE-3, GPGPU, and other accelerators."
On June 30, RIKEN announced that it was taking its five year-old RIKEN Super Combined Cluster System offline so that RICC could take its place. Like the Triton Resource, RICC is also comprised of three distinct systems: a multipurpose parallel PC cluster, a massively parallel PC cluster, and a large capacity memory server. The massively parallel PC cluster has 1,024 of Fujitsu's PRIMERGY RX200 S5 PC servers, with 2,048 CPUs, and 8,192 processing cores. The multipurpose cluster is composed of 100 nodes of Intel Xeon quad-core processors and will be used to run commercial and free software applications, as well as GPU applications. The RICC will use a 550-terabyte hard drive storage system, a tape archive of 2 petabytes and a 10 Gb network environment.
RICC's software stack consists of two Fujitsu solutions: Parallelnavi, an HPC middleware platform and the "Meta-Job Scheduler." The Parallelnavi operating system platform integrates system management functions, high-speed file system, and a program-development environment. The "Meta-Job Scheduler" is an HPC job scheduler tool developed in conjunction with Fujitsu, Fujitsu Laboratories, and RIKEN to maximize efficiency in handling multiple simultaneous jobs and minimize latency.
