Terascale Computing Facility

Terascale Computing Facility
Name	Terascale Computing Facility

Terascale Computing Facility is a supercomputing installation designed for high-performance computation, scientific simulation, and large-scale data analysis. The facility supports computational projects from institutions such as Lawrence Berkeley National Laboratory, Sandia National Laboratories, Argonne National Laboratory, Los Alamos National Laboratory, and Oak Ridge National Laboratory. It provides resources to researchers affiliated with National Science Foundation, Department of Energy, NASA, European Organization for Nuclear Research, and industry partners including IBM, Intel Corporation, and NVIDIA.

Overview

The Terascale Computing Facility delivers teraflop-class and early petaflop-class capabilities for workloads originating at Massachusetts Institute of Technology, Stanford University, University of California, Berkeley, Caltech, and Princeton University. It is used by teams from MIT Lincoln Laboratory, Lawrence Livermore National Laboratory, Center for High Performance Computing, Palo Alto Research Center, and European Commission projects. The user community includes researchers from Harvard University, Yale University, Columbia University, University of Chicago, and University of Michigan. Collaborations span programs supported by National Institutes of Health, Defense Advanced Research Projects Agency, National Aeronautics and Space Administration, European Research Council, and United Nations Educational, Scientific and Cultural Organization.

Architecture and Hardware

The hardware configuration integrates compute nodes sourced from vendors such as Cray Inc., Hewlett Packard Enterprise, Dell Technologies, and Fujitsu with processors from Intel Corporation, Advanced Micro Devices, and accelerators from NVIDIA and AMD (company). The system interconnect uses technologies developed by InfiniBand Trade Association, Mellanox Technologies, and research from Lawrence Livermore National Laboratory networking groups. Storage subsystems combine parallel file systems influenced by designs from The Open Group, such as Lustre (file system), and object storage concepts similar to those used by Amazon Web Services, Google Cloud Platform, and Microsoft Azure. Cooling and power infrastructures incorporate engineering approaches from ASHRAE, American Society of Mechanical Engineers, and best practices trialed at Pacific Northwest National Laboratory and Argonne National Laboratory. The facility supports heterogeneous architectures including multicore CPUs, manycore accelerators, and FPGA prototypes commissioned with assistance from Xilinx, Altera, and Broadcom Inc..

Software and Performance

The software ecosystem runs operating environments based on distributions influenced by Red Hat Enterprise Linux, CentOS, and container technologies from Docker, Inc. and Kubernetes (software). Performance tools include profilers and debuggers such as those developed by Intel Corporation and NVIDIA as well as open-source projects like Open MPI, Message Passing Interface, HDF5, and PETSc. Benchmarking used suites derived from High Performance Linpack, SPEC CPU, and community codes developed at National Energy Research Scientific Computing Center, Los Alamos National Laboratory, and Sandia National Laboratories. Job scheduling and resource management utilize systems inspired by Slurm Workload Manager, Torque (software), and workflow orchestration concepts from Apache Airflow and HTCondor. Performance studies reference work by researchers at Stanford University and University of Illinois Urbana-Champaign.

Applications and Research

Research applications span climate modeling groups at National Oceanic and Atmospheric Administration, astrophysics teams affiliated with Space Telescope Science Institute and Jet Propulsion Laboratory, computational chemistry projects at Scripps Research, and genomics initiatives tied to Broad Institute and Wellcome Sanger Institute. Workloads include simulations for Intergovernmental Panel on Climate Change assessment models, cosmology codes used by European Southern Observatory, and materials science research connected to Max Planck Society and Rutherford Appleton Laboratory. Collaborative engineering projects involve teams from Boeing, General Electric, and Siemens AG. Data-intensive analytics leverage platforms influenced by Apache Hadoop, Apache Spark, and machine learning frameworks originating at Google Research, Facebook AI Research, and DeepMind.

Facility Operations and Management

Operations are overseen by administrative structures modeled after institutional practices at Lawrence Berkeley National Laboratory and Oak Ridge National Laboratory, with governance interaction involving National Science Foundation program officers and Department of Energy program managers. User support and training programs draw on curricula developed at Argonne National Laboratory and NERSC user engagement efforts. Security and compliance integrate standards recommended by National Institute of Standards and Technology and collaborations with Cybersecurity and Infrastructure Security Agency. Procurement and lifecycle planning reference contract frameworks used by General Services Administration and procurement case studies from University of Cambridge and Imperial College London.

History and Development

The facility originated from initiatives funded by National Science Foundation grants and Department of Energy allocations, shaped by early-generation systems at Los Alamos National Laboratory and Sandia National Laboratories. Design and construction referenced architectures demonstrated at National Center for Supercomputing Applications and Pittsburgh Supercomputing Center, with procurement influenced by market offerings from Cray Inc. and IBM. Research milestones involved collaborations with Oak Ridge National Laboratory and publication venues such as ACM SIGARCH, IEEE Computer Society, and Journal of Computational Physics. Ongoing upgrades have incorporated technologies from Intel Xeon Phi, NVIDIA Tesla, and developments in exascale planning discussed at International Conference for High Performance Computing, Networking, Storage and Analysis.

Category:Supercomputing facilities