Supercomputers

Supercomputers Supercomputers are high-performance computing systems built for extreme computational tasks, delivering orders of magnitude greater processing power than mainstream Intel Corporation, Advanced Micro Devices, NVIDIA Corporation, or ARM Holdings based servers. They are developed and operated by institutions such as Lawrence Livermore National Laboratory, Los Alamos National Laboratory, Oak Ridge National Laboratory, CERN, and commercial providers like IBM, Hewlett Packard Enterprise, and Cray Inc. used by projects including the Human Genome Project, Large Hadron Collider, and climate initiatives from the Intergovernmental Panel on Climate Change. The design, deployment, and operation intersect communities around procurement by agencies such as the United States Department of Energy, collaborations like EuroHPC Joint Undertaking, and national programs in China, Japan, European Union, India, and Russia.

History

Early milestones trace to machines developed by entities such as Los Alamos National Laboratory and companies like Cray Research and Control Data Corporation that built vector processors and early parallel systems for projects like Manhattan Project-era scientific computation and Cold War-era simulations tied to work by the United States Department of Energy and NATO partners. Landmark systems appeared from manufacturers including CDC 6600 designers and successors like Cray-1; later generations involved companies such as IBM with systems used for initiatives like Blue Gene and collaborations with national labs including Argonne National Laboratory. National competitions and lists such as the TOP500 and awards like the Gordon Bell Prize chronicle shifts from vector architectures to massively parallel clusters influenced by processors from Intel Corporation, accelerators from NVIDIA Corporation, and interconnects from firms like Mellanox Technologies. International programs in Japan produced systems used by organizations like RIKEN, while Chinese efforts at institutions such as National University of Defense Technology and projects involving Tianhe-2 and Sunway TaihuLight reflect strategic investments by governments including those of People's Republic of China and collaborations with ministries akin to Ministry of Science and Technology (China).

Architecture and Design

Designs combine components from suppliers such as Intel Corporation, Advanced Micro Devices, NVIDIA Corporation, and custom processors like those from Fujitsu or the Sunway project at National University of Defense Technology. Architects balance compute nodes, memory subsystems provided by vendors like Micron Technology and Samsung Electronics, and storage systems from companies such as NetApp and Seagate Technology. Network topologies employ interconnects and switching technologies from firms including Mellanox Technologies and Cisco Systems, using designs like fat-tree, torus, and dragonfly used in systems at laboratories such as Lawrence Berkeley National Laboratory. System software and firmware originate from projects tied to Linux Foundation distributions, custom kernels developed by IBM or Hewlett Packard Enterprise, and orchestration by management frameworks from organizations like OpenStack Foundation and standards bodies such as OpenMPI Project.

Performance and Benchmarking

Performance measurement relies on benchmarks and lists maintained by groups including the TOP500 project, the High Performance Linpack (HPL) benchmark, and prizes like the Gordon Bell Prize. Metrics such as FLOPS are used in comparisons applied to systems at institutions like Argonne National Laboratory and international contenders from Japan and China. Alternate benchmarks and suites developed by communities such as the SPEC consortium, the Graph500 initiative, and the Green500 rank systems for different workloads and energy efficiency criteria tracked by organizations including Lawrence Berkeley National Laboratory.

Applications

High-performance systems serve research centers like CERN for particle physics, national laboratories such as Oak Ridge National Laboratory for materials science, and universities including Massachusetts Institute of Technology and Stanford University for computational chemistry and climate modeling used by organizations including the World Meteorological Organization and projects like the Coupled Model Intercomparison Project. They underpin simulations for aerospace work by companies such as Boeing and research at agencies like NASA, genomic analysis from consortia following the Human Genome Project, financial risk modeling in firms comparable to Goldman Sachs or JPMorgan Chase, and machine learning training by enterprises including Google, Microsoft, and Facebook. Supercomputing resources often support initiatives coordinated by consortia such as PRACE in the European Union and national allocations via bodies like the National Science Foundation.

Software and Programming Models

Software ecosystems include operating systems and middleware from projects like Linux Foundation distributions and libraries from the OpenMPI Project and programming standards such as MPI and OpenMP used by research groups at institutions like Los Alamos National Laboratory and Argonne National Laboratory. Programming languages and frameworks from organizations such as NVIDIA Corporation (CUDA), communities around OpenACC, and vendor tools from Intel Corporation and AMD complement scientific libraries like those from the Netlib repository and numerical software used by teams at Massachusetts Institute of Technology and Stanford University. Development, debugging, and workflow orchestration involve platforms and projects like GitHub, Jenkins, and data management stacks from companies such as IBM and Dell Technologies.

Energy Efficiency and Cooling

Energy management and cooling strategies are engineered by collaborations among vendors like Schneider Electric and research programs at laboratories such as Lawrence Berkeley National Laboratory. Approaches include liquid immersion pioneered by labs and companies including 3M partnerships, direct liquid cooling used in deployments by Fujitsu and Hewlett Packard Enterprise, and data center designs influenced by standards bodies such as ASHRAE. Rankings like the Green500 highlight energy-efficiency performance of installations at facilities run by institutions such as Oak Ridge National Laboratory and national centers in Japan and China.

Future Trends and Challenges

Future directions engage exascale initiatives led by programs such as the Exascale Computing Project and multinational efforts including EuroHPC Joint Undertaking, with hardware roadmaps from firms like Intel Corporation, AMD, NVIDIA Corporation, and Fujitsu. Challenges span supply chains involving companies such as TSMC and research into heterogeneous architectures pursued at universities including Massachusetts Institute of Technology, University of California, Berkeley, and national labs like Argonne National Laboratory. Algorithmic and software scalability concerns are addressed by communities around the Gordon Bell Prize, while geopolitical and policy implications involve agencies such as the United States Department of Energy and international collaborations like CERN and PRACE.

Category:Computing