Distributed Analysis of Neutron Scattering Experiments on the TeraGrid

Neutron scattering research in the United States is poised for a spectacular leap after the Spallation Neutron Source (SNS) is commissioned in 2006. With the power of its source and the performance of its instruments, the SNS will improve measurement efficiencies by typically two orders of magnitude. Analyzing data is an essential but non-trivial step in neutron scattering research. The productivity of the SNS for cutting-edge science will be enhanced considerably by modern computing. It is in the national interest for the SNS to deliver quickly on its promise of scientific advances in our understanding of the structure and dynamics of materials and condensed matter, and the sophistication of its scientific results will be enhanced by the sophistication of its computing resources. The past decade has seen a spectacular leap in the capabilities of computational science. Dramatic improvements in computing hardware have been matched by dramatic improvements in our understanding of how to compute the properties of materials, molecules, and condensed matter. Many methods, once exotic, have become routine with improvements in computing hardware and software. These methods, such as those of quantum chemistry and condensed matter science, can address many of the same phenomena measured experimentally by neutron scattering research. A number of new scientific opportunities, some discussed below, involve a closer interaction between computational science and neutron scattering science. In the near future, these new directions will therefore track the trends in computing resources that are followed by computational science, although experimental neutron science brings a number of new challenges such as near real-time computing. An important long-term trend, still in its infancy, is distributed computing, which offers new flexibility and power of computing resources. Over the past several months, we have considered a number of ways where distributed computing could be utilized for the support of neutron science. In particular, we have considered how the evolving capabilities of the pyre system that underlies DANSE (distributed data analysis for neutron scattering experiments), the TeraGrid project funded by the NSF, and a neutron science portal at the SNS could offer new capabilities for neutron science, the types of interactions between these systems that would be plausible and useful, and what opportunities exist for improved designs. These issues are addressed in the present report.