Earth Simulator

Earth Simulator
Name	Earth Simulator
Location	Yokohama, Japan
Operator	Japan Agency for Marine-Earth Science and Technology
Introduced	2002
Peak performance	40 PFLOPS (post-upgrade)
Architecture	Vector processing, distributed memory
Purpose	Climate simulation, geophysics, environmental modeling

Earth Simulator The Earth Simulator is a high-performance supercomputer system deployed in Yokohama and operated by the Japan Agency for Marine-Earth Science and Technology, designed for large-scale simulation of climate change, seismology, oceanography, and related geophysics problems. It has driven research collaborations among institutions such as the University of Tokyo, National Institute for Environmental Studies, JAXA, and international partners, enabling work tied to events like the 1990s climate modeling expansion and projects funded by agencies including the Ministry of Education, Culture, Sports, Science and Technology (Japan).

Overview

The system was commissioned to address computational demands in studies of global warming, paleoclimate reconstruction, earthquake dynamics, and tsunami propagation, providing dedicated resources for initiatives led by bodies such as the Intergovernmental Panel on Climate Change and the World Climate Research Programme. It supported campaigns coordinated with research centers including the Geological Survey of Japan, the National Oceanic and Atmospheric Administration, and the European Centre for Medium-Range Weather Forecasts, facilitating analysis at resolutions previously unattainable by earlier machines like the Cray-1 and contemporaneous systems such as the ASCI White.

Architecture and Hardware

The original hardware employed vector processors inspired by designs from companies like Fujitsu and utilized high-bandwidth interconnect topologies similar to those in systems from NEC and Hitachi. Nodes incorporated large shared-memory vector units and distributed memory fabrics influenced by designs seen in the IBM Blue Gene line and later compared with Fujitsu K computer configurations. Storage subsystems interfaced with parallel file systems used in centers like Lawrence Livermore National Laboratory and networking stacks compatible with infrastructure at the National Institute for Informatics. Cooling and power delivery drew on facility engineering practices from installations at Oak Ridge National Laboratory and Argonne National Laboratory.

Software and Modeling Techniques

Software stacks included optimized numerical libraries, compilers, and communication layers developed in collaboration with vendors such as Fujitsu and research groups at the University of Tokyo and Kyoto University. Modeling frameworks ran components for atmospheric models similar to those in ECMWF workflows, ocean models akin to those developed by the National Oceanography Centre, and geodynamic simulators used by the US Geological Survey. Techniques integrated high-resolution finite-difference, spectral element, and finite-volume schemes established in literature associated with institutions like MIT, Princeton University, and Caltech to simulate interactions among the atmosphere, ocean, and solid Earth.

Performance and Benchmarking

At launch, peak performance records were compared against the TOP500 list and metrics from machines such as Earth Simulator 2 successors and predecessors like NEC SX-9 and the Cray XT5. Benchmarking used standardized tests including LINPACK and application-specific proxies derived from climate codes used at NOAA and seismology benchmarks from the USGS. Reported sustained throughput on production workloads influenced procurement decisions at agencies such as the Ministry of Economy, Trade and Industry (Japan) and research centers like the Riken Center for Computational Science.

Scientific Applications

Researchers employed the platform to run coupled atmosphere-ocean models addressing phenomena like El Niño–Southern Oscillation, Pacific Decadal Oscillation, and monsoon variability studied at the International Pacific Research Center. Tsunami modeling tied to the 2011 Tōhoku earthquake and tsunami enabled hazard assessments used by the Cabinet Office (Japan). Paleoclimate simulations informed assessments by the Intergovernmental Panel on Climate Change, and seismic wave propagation studies contributed to understanding rupture mechanics in regions monitored by the Japan Meteorological Agency and the Seismological Society of Japan.

Development History and Upgrades

Initial deployment in the early 2000s followed planning phases involving universities including Tohoku University and national labs such as JAMSTEC partners. Subsequent upgrade cycles incorporated lessons from large-scale projects at Sandia National Laboratories and benchmarking experiences with machines like K computer and later exascale prototypes. Funding and oversight engaged ministries such as the Ministry of Education, Culture, Sports, Science and Technology (Japan) and collaborations with international consortia that included researchers from NOAA, ECMWF, and the Australian CSIRO.

Criticisms and Controversies

Critiques focused on cost-benefit analyses debated by stakeholders including policy bodies such as the Diet of Japan and academic commentators from institutions like Keio University and Waseda University, raising questions about allocation of capital toward centralized supercomputing versus distributed computing resources advocated by groups such as OpenScience proponents. Concerns about reproducibility and software portability echoed discussions occurring at forums like the Supercomputing Conference and in publications by researchers affiliated with Imperial College London and ETH Zurich.

Category:Supercomputers Category:Climate modeling Category:Computational geophysics