Modelling for the Chemistry using Industries
Smith Institute: Modelling for Chemistry: Industry Day, Runcorn, 27 November 2007
Grid and High Performance Computer Modelling for the Chemistry-Using Industries.
In July this year the Chemistry Innovation KTN, after a wide consultation with industry, launched the first of what will be regular meetings aimed at identifying the necessary actions required to improve the impact of modelling on innovation in the chemistry-using industries. Key to this is correct identification of the technology challenges industry faces in applying modelling tools and addressing how these might be overcome. One clear message from the meeting was that industry was not well informed on what R and D is ongoing in the field and whether or not it will prove to be helpful in solving some very real problems.
In other words they asked "what are academia and the computer modelling industry doing for us?"
Consequently, Daresbury Laboratory in conjunction with Chemistry Innovation is organising a day-long Workshop designed to answer this question by achieving these objectives:
- Showcase some current R and D activities that are ongoing in both model development and computer performance improvement
Identify any areas that continue to cause difficulties that are not being addressed and so catalyse the R&D necessary to solve those problems.
- Introduce GRID and High Performance Computing (HPC) and illustrate some of the ways in which they can overcome otherwise intractable modelling problems.
The Problems:
Computer modelling is now used in practically all areas of industry, commerce and academia. It is possible to carry out, in the virtual world of a computer, huge numbers of virtual experiments that would, in the real world, entail vast expenditure of time, money and effort. These virtual experiments help to focus scarce resources onto the most promising areas and allow real experiments to be carried out with the greatest chances of success. However the more exact and large the virtual experiment, the greater the computer power needed to complete that experiment in a reasonable time. Traditionally this problem has steadily diminished thanks to the fact that computers have rapidly become ever more powerful. This is no longer the case. Also the very algorithms and models available for use in virtual experimentation do not cover all situations and circumstances.
In short the value of computer modelling is being undermined due to a perceived lack of good enough models and of fast enough computers.
Some Solutions:
During the Workshop we will show how work at STFC's Daresbury Laboratory, home of its Computational Science and Engineering Department, and elsewhere is trying to solve these problems by attacking them on three fronts;
- developing algorithms, codes and models to enable the most realistic possible of virtual experimentation to be carried out
- optimising the operation of codes and software to take
- adapting code and software to run in parallel on multiple (say 10,000) processors simultaneously
Also the "NW-GRID" is at Daresbury Lab, which is a GRID computer with over 300 processors at Daresbury and over 600 more located at the Universities of Lancaster, Liverpool and Manchester. With over 1000 connected processors this is a true Supercomputer and it is available for use by commerce and industry courtesy of the North West Development Agency, its primary sponsor. The Workshop will introduce NW-GRID and illustrate its capabilities and how to access them. Supercomputing on a "Pay per Use" basis!
Further information about the "Modelling for Chemistry" priority area within Chemistry Innovation and how you can get involved is available from Maureen Laughton, Heather Tewkesbury or visit www.chemistryinnovation.co.uk.