HECToR national supercomputing service
HECToR (High End Computing Terascale Resource) was the national high-performance computing service for the UK academic community from 2008–2014.
It gave UK scientists the means to undertake complex computational simulations across a range of scientific disciplines including climatology, earth sciences, chemistry, materials, fluid dynamics, atomic and molecular physics, plasma physics and nanoscience. The service, which began in 2008 and operated for six years, had a final theoretical peak capacity of over 800 Tflop/s and 90,112 processing cores.
The computer hardware was provided by Cray and it was accommodated and managed by UoE HPCX Ltd at the University of Edinburgh's Advanced Computing Facility.
The HECToR service was replaced by ARCHER in 2014.
Enhancing science on HECToR
EPCC works closely with researchers to enhance the performance of key scientific and engineering applications and enable them to optimally exploit large-scale resources such as HECToR. Four examples are given here.
EPCC, in collaboration with Division of Pathway Medicine at The University of Edinburgh, developed SPRINT, which provides easy access to high-performance computing for the analysis of high throughput post-genomic data using the statistical programming language R. This work involved porting to, and specific optimisation for, HECToR.
Computational Fluid Dynamics
The scalability of EBL, a code designed to help answer fundamental research questions in the area of turbulence, was been dramatically improved by EPCC from hundreds of cores to at least 14,000 cores
These improvements were achieved by restructuring to implement a two-dimensional "pencil" domain decomposition.
EPCC performed dramatic optimisations on CARP, a cardiac simulation code, taking the application a step closer to clinical use. The simulation of a human heartbeat, which previously took over an hour, is now possible in under 5 minutes. These improvements involved development of a parallel mesh decomposition scheme and asynchronous parallel output.
EPCC continues to actively develop and improve the popular CP2K and VASP materials chemistry applications. Improvements, which have resulted in impressive performance gains, include optimisation of communication patterns, restructuring to improve load balance, and introduction of mixed mode programming models to allow more effective utilisation of multicore systems.
- Harnessing Thousands of Cores Using CP2K
- Speeding-up Simulation Codes on HECToR