HECToR

HECToR
Name	HECToR
Country	United Kingdom
Operator	Engineering and Physical Sciences Research Council
Location	Edinburgh
Manufacturer	Cray Inc.
Date operational	2007
Date decommissioned	2014

HECToR HECToR was a high-performance computing service hosted in the United Kingdom, designed to provide computational capability to researchers in physical sciences, engineering, climate science, and computational chemistry. It served universities, national laboratories, and industrial partners, linking academic users with resources comparable to those deployed at Argonne National Laboratory, Oak Ridge National Laboratory, and Lawrence Berkeley National Laboratory. Funded and overseen by national and regional bodies, HECToR formed part of the European and global high-performance computing ecosystem alongside systems such as Jaguar (supercomputer), Blue Gene/L, and Cray XT4.

Overview

The service provided a cluster-scale platform aimed at enabling large-scale simulations in areas including fluid dynamics, materials modelling, astrophysics, and climate modelling. HECToR integrated compute hardware, interconnect topologies, storage subsystems, and software stacks similar to those used on Tianhe-1, Fujitsu K Computer, and Sequoia (supercomputer). As an instrument of the Engineering and Physical Sciences Research Council and partner institutions, HECToR facilitated collaborations with international projects funded by bodies such as the European Commission and research initiatives with groups like Met Office scientists and teams from University of Cambridge, University of Oxford, and Imperial College London.

History and Development

The HECToR project originated from strategic funding allocations by UK research councils to establish national capability for computational science. Early planning involved procurement processes akin to those conducted by National Centre for Atmospheric Research and procurement consortia such as those supporting Cori (supercomputer). Initial installation phases engaged vendors including Cray Inc. and system integrators familiar with deployments at Los Alamos National Laboratory and Sandia National Laboratories. Over successive upgrade cycles, the service mirrored transitions seen in systems at European Centre for Medium-Range Weather Forecasts and hardware refreshes comparable to upgrades at MareNostrum and Piz Daint.

HECToR’s operational governance involved academic consortia, regional computing centres, and national funding agencies; its user base encompassed principal investigators from institutions such as University of Edinburgh, University of Manchester, and University of Southampton. Periodic review panels composed of experts from STFC and panels similar to those formed by Royal Society provided strategic oversight, aligning HECToR with national research strategies and industrial engagement initiatives drawing parallels with collaborations at Siemens and Rolls-Royce.

Architecture and Hardware

The platform comprised multi-core compute nodes, memory hierarchies, and high-performance interconnects reflecting architectural trends from contemporaneous systems like Cray XT4, IBM Blue Gene/Q, and Sunway TaihuLight. Processor technologies evolved during its lifecycle, transitioning through CPU families comparable to those used in systems at CEA and Riken. Interconnects adopted topologies and link technologies similar to those implemented at Cisco Systems and Infiniband Trade Association member deployments, while storage subsystems used parallel file system designs akin to Lustre and implementations at NERSC.

Cooling and power infrastructure reflected standards adopted by data centres hosting Facebook (company) research clusters and national e-infrastructure sites such as Hartree Centre. System management incorporated batch schedulers and resource managers comparable to software used at Lawrence Livermore National Laboratory and Argonne National Laboratory to orchestrate job queues and optimize utilization.

Performance and Benchmarks

HECToR achieved performance metrics reported in system-level benchmarks and application-specific scaling studies, with throughput and floating-point rates evaluated using suites similar to LINPACK and community benchmarks employed at Top500 listings. Comparative analyses were performed versus contemporaneous platforms like Jaguar (supercomputer), Tianhe-1, and campus clusters at University College London and ETH Zurich. Users published scaling results in venues such as proceedings of SC (conference), ISC High Performance, and journals affiliated with Institute of Physics and American Physical Society.

Application benchmarks for computational fluid dynamics, climate models, and materials science used codes and frameworks comparable to WRF, CESM, LAMMPS, and Quantum ESPRESSO, with performance tuning drawing on libraries like those developed by Intel Corporation and optimizations pioneered in collaborations with groups at Argonne National Laboratory and Oak Ridge National Laboratory.

Software Environment and Applications

The software stack comprised compilers, numerical libraries, parallel programming models, and workflow tools analogous to ecosystems at NERSC and PRACE centres. Users accessed environments supporting MPI implementations used at National Institute for Computational Sciences, shared libraries including those from Netlib, and domain-specific packages common at CERN and Max Planck Institute for Plasma Physics. Scientific applications targeted by users included climate simulation codes used by Met Office Hadley Centre, quantum chemistry codes employed by teams at University of Cambridge, and engineering solvers utilized by researchers affiliated with Rolls-Royce and BAE Systems.

Training and user support mirrored programs run by XSEDE and national supercomputing services, with workshops and documentation produced in partnership with academic partners such as University of Edinburgh and Heriot-Watt University.

Decommissioning and Legacy

HECToR was retired following planned replacement cycles, with decommissioning activities coordinated among funding bodies and successor infrastructures similar to transitions to systems like ARCHER and European petascale platforms funded through European Union initiatives. Legacy outcomes included datasets, software optimizations, and trained cohorts of computational scientists who moved into projects at STFC, EPSRC-funded consortia, and industrial research groups such as Schlumberger and BAE Systems. Lessons from HECToR informed procurement, architecture selection, and support models adopted by later UK facilities and contributed to collaborations with international centres including PRACE and national labs like Oak Ridge National Laboratory.

Category:Supercomputers