All ENACTS partners are working together for the ultimaate goal, to foster the fast and efficient uptake of the latest HPC technologies, and move us towards our goal of European Grid computing, a "virtual infrastructure" where each researcher, regardless of nationality or geographical location, has access to the best resources and can conduct collaborative research with top quality scientific and technological support.
In ENACTS, our strategy involves close co-operation at a pan-European level, to review service provision and distill best practice, to monitor users' changing requirements for value-added services, and to track technological advances in a number of specific areas. Since 2000, two extensive studies have been performed: Grid Service Requirements and HPC Technology Roadmap. The results and recommendations are published in reports that are freely available to all Grid community at all levels and have already received very positive feedback.
"Sun Microsystems is strategically committed to the Grid and views the information contained in the ENACTS report as an important addition to its understanding of the HPC Grid market", state John Beales, Business Technologist, at Sun Microsystems.
The reports have also been included into the reading list for the MSc in High Performance Computing that EPCC at the University of Edinburgh is running. Dr. David Henty, Training and Research Manager at EPCC, stated: "The ENACTS reports will help the students studying for the MSc in High Performance Computing to get a good grasp of present and future issues that the Grid is facing. It is very important as today's students will be the next generation of Computational Scientists who will have to resolve these issues."
With a few exceptions the actual beneficiaries and users of tomorrow's Grid technology have not yet established a dialogue with the people involved in the standartisation and development of the Grid middleware and the implementation of higher-level services. Their requirements are thus often at best assumed or second-guessed, or simply ignored in the worst case. Clearly, a rapid acceptance and the following uptake of such a technology in the user community will be key to its success.
Within the ENACTS framework, EPCC at the University of Edinburgh in the United Kingdom, in collaboration with the Poznan Supercomputing and Networking Centre (PSNC) in Poland have undertaken a review of Grid models and their adequacy for various user groups and application types. These results are based on a survey of 85 European group leaders issued from mainstream computational science disciplines.
This study has produced an extensive report. Terry Sloan, Project Manager at EPCC, stated: "The Grid Service Requirements report has proven to be of great use to the Sun Data and Computing Grids project, which is a DTI/EPSRC-funded collaboration between Sun Microsystems and EPCC. It has helped us assess the scientific community's expectations from Grid technologies and more importantly aided us in the formulation of our project requirements."
The report contains detailed description of the different types of computational Grids, with a discussion of their features, limitations, and FAQ for prospective users. This section is of particular interest to operators of HPC systems and systems administrators who want to find out more about Grid computing.
Results from a comprehensive survey undertaken among 85 major European research groups have also been published in the report. This covered aspects such as awareness and expectations of Grid computing, groups and applications profile, working practices, bottlenecks and requirements, future needs, etc. The purpose of the survey is to gain a better understanding of these key user
groups' composition and working practices in view of establishing their key requirements.
The final section of the report provides conclusions and recommendations. This section summarises the features offered by the leading-edge Grid middleware, compares this to the users' requirements, and following these, makes recommendations on a possible strategy for a successful uptake of Grid technology within the user community.
The HPC Technology Roadmap study, performed jointly by National Supercomputing Centre in Linkoping, Sweden and CNR Center for High Performance Computing in Molecular Sciences at the University of Perugia, Italy, investigates technological and economic trends in the development of HPC hardware architectures and systems in 2006-2011. A case study for the area of Molecular Modelling has also been performed and included in the report.
The information has been obtained by means of interviews with representatives of seven vendors of HPC systems including Compaq, Sgi, Sun, IBM, Cray, NEC, and Fujitsu. Vendors are often portrayed as having quite different views on supercomputing, especially in the areas such as processors, vectorisation, memory architecture and programming models. Interestingly enough, the interviews showed that there is general agreement on many of the major trends in the future computer architectures in general and HPC in particular.
Generally speaking, the computing systems of the future can be described as scalable parallel architectures based on the notion of clustered SMPs. When looking "under the hood" however, there is a wide array of different solutions presented.
New advances in all relevant computer technologies, made possible by an increasing consumer market as well as the Open source movement, seems to make the number of available solutions even larger. And if anyone believed that future HPC systems would all move towards a homogenous structure, it seems that they will be disappointed.
The study concludes with the case study for the key molecular science community, performed by experts on molecular applications at CSCISM in Perugia, Italy. Using the results from the vendor interviews, two typical HPC solutions are configured. This is used when investigating the impact of future HPC development on three main approaches to the definition of interatomic potential:
- Model potential based on Force-Field (FF) parametrisations,
- DFT (Density Functional Theory) derived otentials, and
- Ab initio potentials.