This article was accepted into the corpus but its outbound wikilinks were never NER-processed — typical at the deepest BFS hop or when the run's entity cap was reached. No expansion funnel to show.
| Eddington (development) | |
|---|---|
| Name | Eddington |
| Developer | Theoretical Cosmology Consortium |
| Manufacturer | Advanced Research Computing Ltd. |
| Release | 2013 |
| Type | Cosmology supercomputer |
| Cpu | Intel Xeon |
| Gpus | NVIDIA Tesla |
| Os | Linux |
| Memory | 1 PB |
| Storage | 10 PB |
| Location | Institute of Cosmological Computing |
Eddington (development)
Eddington (development) was a dedicated cosmology-focused supercomputing project initiated to support large-scale simulations and data analysis for precision cosmology. The project involved collaborations among major research institutes, high-performance computing vendors, and funding bodies to deliver a platform for numerical relativity, N-body simulations, and Bayesian inference. Its aims connected theoretical work, observational programs, and software ecosystems across astrophysics, particle cosmology, and computational science.
The Eddington initiative grew from discussions at Royal Astronomical Society workshops and planning meetings between groups at University of Cambridge, California Institute of Technology, Princeton University, Max Planck Society, and Lawrence Berkeley National Laboratory. Early feasibility studies referenced simulation campaigns by teams at Los Alamos National Laboratory, Argonne National Laboratory, and Oak Ridge National Laboratory, and drew on designs from the DiRAC and PRACE infrastructures. Funding proposals were submitted to agencies including European Research Council, National Science Foundation, Science and Technology Facilities Council, and NASA; hardware procurements involved bids from IBM, Dell EMC, NVIDIA Corporation, and Intel Corporation. The project timeline intersected with survey missions and facilities such as Planck (spacecraft), Sloan Digital Sky Survey, Dark Energy Survey, Large Synoptic Survey Telescope, and theoretical advances from groups at Institute for Advanced Study and CERN. Governance employed advisory input from panels of researchers affiliated with Cambridge University, Harvard University, Stanford University, and ETH Zurich.
Eddington's architecture was co-designed by teams from Cray Inc., Hewlett Packard Enterprise, and university HPC centers including Pawsey Supercomputing Centre and NVIDIA Research. The system architecture balanced compute hierarchies inspired by designs at FLOPS-centric installations such as Blue Gene/Q and hybrid GPU-accelerated deployments seen at Titan (supercomputer). Network topology choices reflected lessons from InfiniBand fabric deployments at Oak Ridge Leadership Computing Facility and routing strategies used by NERSC and EPCC. File system and storage strategies borrowed concepts implemented at Storage Resource Broker projects and large-scale arrays used by European Space Agency archives. Cooling and power design consulted with teams behind Green500 entries and datacenter architects from Microsoft Research and Google DeepMind.
The hardware stack featured multi-socket Intel Xeon processors and NVIDIA Tesla GPUs, mirroring configurations used at Fermi National Accelerator Laboratory and Lawrence Livermore National Laboratory. Performance benchmarking referenced LINPACK runs comparable to submissions from TOP500 systems and mixed-precision workloads emulating efforts at NVIDIA Volta deployments. Memory hierarchies and NUMA layouts were informed by studies from AMD Research and architectural analyses published by ACM authors linked to Stanford Linear Accelerator Center. I/O throughput goals targeted parity with high-throughput archives used by European Organisation for Nuclear Research and data rates consistent with pipelines developed for Square Kilometre Array. Reliability engineering was influenced by fault-tolerance work from NASA Ames Research Center and checkpointing research from Sandia National Laboratories.
Eddington supported scientific codes and toolchains used by teams at Max Planck Institute for Astrophysics, Kavli Institute for Cosmology, and Perimeter Institute. Simulation packages ported included codes analogous to GADGET, ENZO, RAMSES, and lattice tools used at CERN Theory Department. Analysis and statistical frameworks integrated libraries and workflows similar to NumPy, SciPy, TensorFlow, and probabilistic tools like those developed at Alan Turing Institute. Job scheduling and resource management resembled systems from SLURM Workload Manager and orchestration methods practiced at Argonne and NERSC. Visualization pipelines paralleled environments used at Jet Propulsion Laboratory and Visualization Sciences Group to render outputs for teams at Space Telescope Science Institute.
Operations were coordinated through an operations center modeled on practices at European Southern Observatory and national centers such as PRACE and XSEDE. Deployment phases aligned with commissioning tests similar to those used at Large Hadron Collider experiments, with acceptance criteria involving research groups from Institute of Cosmology and Gravitation and computational scientists from University of Edinburgh. Security and data governance drew on protocols from UK Research and Innovation and international agreements involving CODATA-affiliated bodies. User support, training, and documentation efforts paralleled outreach from Software Carpentry and curriculum initiatives at Imperial College London and University College London.
Eddington influenced simulation capabilities at institutions including National Institute for Computational Sciences, Swiss National Supercomputing Centre, and research groups at University of Tokyo and Seoul National University. Publications arising from Eddington-enabled projects appeared in journals such as Monthly Notices of the Royal Astronomical Society, The Astrophysical Journal, Physical Review D, and Nature Astronomy and were cited by collaborations behind Euclid (spacecraft), WFIRST and ground-based programs like Subaru Telescope. The platform informed procurement and design decisions for later systems at Oak Ridge National Laboratory and influenced training programs at CERN School of Computing. Broader impacts included methodological advances used by teams at IBM Research and algorithmic improvements adopted by groups at Google AI and Microsoft Research.