About the Miracle Consortium
The Miracle Consortium (henceforth Miracle) was originally set up as part of the Particle Physics and Astronomy Research Council's (PPARC) 1998 High Performance Computing initiative and comprised of astronomers from University College London, University of Hertfordshire, Imperial College London and Queen's University Belfast.
A brief history of the consortium is provided below, with key dates highlighted.
In 1999, Miracle entered a collaboration with Chemistry, Physics, Earth Science (all UCL) and the computer company SGI to build a local high performance computing facility in the physical sciences. This facility was called HiPerSPACE (or High Performance Service for Physics, Astronomy, Chemistry and Earth Sciences) and consisted of two large parallel computers based on SGI's highly successful Origin 2000 machines and managed by Callum Wright (now at the University of Bristol). Of the two machines, the Miracle astronomers had dedicated access to the one called MIRACLE (Multi-Institution Research Astrophysical Computing) which had 24 Origin 2000 processors and 6 GB of RAM; the other machine was named Bentham after the philosopher who inspired the foundation of UCL.
|The HiPerSPACE facility was opened in November 1999 by the Minister of Science at the time Lord Sainsbury. At peak performance the two machines of the HiPerSPACE facility could carry out between 45 and 50 billion individual calculations per second which meant that at the time of installation, it was the UK's most powerful computer dedicated to the physical sciences.||
Press Release Highlights
- Earth Impacting Sun Storms Predicted (from PPARC) (November 1999)
Between 2001 and 2002, Miracle augmented its computational capability with the purchase of the Enigma and Parameter Search Engine (PSE) computers which were based on Sun Microsystem's V880 servers (based on the Ultrasparc III processor) and were managed by Callum Wright as part of the HiPerSPACE facility.
The Enigma computer was funded under the PPARC Joint Research Equipment Initiative (JREI) in early 2001. It had 16 processors, 64GB of RAM and a fast Myrinet connection to enable complex parallel simulations to be performed.
The Parameter Search Engine was funded under a PPARC Theory Research Assessment Panel (TRAP) grant in early 2002. It had 24 processors and 96GB of RAM and was used to perform large parameter searches and library calculations.
Below are some highlights of the work conducted on these machines.
- The ab initio calculation of the spectra of atomic ions to very high ionisation states and energies (Reid and Bell QUB)
- The ab initio calculation of the spectra of H2O up to very high rhovibrational energy states (Tennyson UCL).
- The ab initio calculation of the optical properties of astrophysical dust particles and polymers such as poly-aromatic-hydrocarbons (Gledhill UH, Gray UMIST, Yates UCL).
- The numerical simulation of the dynamics and chemistry of planetary atmospheres (Aylward and Miller UCL, Muller IC).
- The numerical simulation of the physical conditions in ionised regions in the Local Universe (Barlow, Storey and Ercolano UCL).
- The numerical simulation of the chemical environments in star formation regions including low mass protostars and high mass hot cores (Viti and Rawlings UCL, Gray and Millar UMIST).
- The calculation of detailed radiative transfer in dusty and molecular environments such as star formation regions, evolved stars, starburst regions and active galactic nucleii (Yates, Bowey and Ercolano UCL, Gledhill and Lucas UH, Gray UMIST).
- The calculation of stellar structure and emission properties of massive stars,including the effects of rotation and metallicity (Howarth, Smith, Prinja, Townsend and Burnley UCL).
- The calculation of key cosmological parameters (Lahav and Bridle UCL).
The increasing complexity and resolution of observational data was requiring theorists to explorer greater volumes of parameter space at finer resolutions to make observable predictions and to calculate the underlying physics, chemistry and physical conditions that created the phenomena being observed. To cope with this increasing demand for computational power the Miracle Consortium, which now included astronomers from the University of Strathclyde and the University of Manchester, received a grant from PPARC to upgrade computational ability.
The consortium engaged in a joint venture with UCL Research Computing to upgrade the existing HiPerSPACE service resulting in the development of the 224 processor Keter cluster. Keter was based on a mixture of Sun V880 (each containing 8 UltraSparc III processors and between 16 and 32GB of RAM) and Sun V890 servers (each containing 8 Ultrasparc IV processors and 64GB RAM) connected over a high performance 5Gbps Myrinet network. The Miracle Consortium was provided with guaranteed primary access to 50% of Keter enabling the astronomers to, for example, create the largest ever water line list and discover an unexpected cooling effect in Saturn's atmosphere.
In 2010, the Miracle Consortium was awarded funding to purchase CPU time from UCL Research Computing's Legion High Performance Computing Cluster; Legion is a distributed memory computing cluster with 2560 cores, 4GB RAM per core and Infinipath chip-to-chip connectivity.
Page last modified on 11 feb 11 14:31 by Dugan K Witherick