Automation

 

A natural consequence of the effective utilization of supercomputers is the tremendous amount of data generated at remote locations, e.g. where supercomputing centres are located, presenting data management and data grid technology challenges. The original data must be efficiently analysed to compute derived quantities. New advanced visualization techniques are needed to help identify key features in the data. There are also significant programming and algorithmic challenges, which must be met in order to enable computational capabilities for addressing more complex scientific problems. These include multi-dimensional domain decomposition in toroidal geometry and mixed distributed/shared memory programming. Other problems include load balancing on computers with large numbers of processors, optimization of fundamental gather-scatter operation in particle-in-cell codes and scalable parallel input/output operations for the petascale range of data sets.

Another key priority to help accelerate progress on the impressive state-of-the-art physics advances involves developing a set of diagnostic and visualization tools that will allow real-time interaction with the simulated data. This will impact the ability of scientists to effectively test theories/hypotheses and to address specific questions about the proper implementation of the key physics within the computational models. Recent efforts using GKV, an interactive data analysis and visualization tool expressly designed for analysing output from plasma microturbulence simulations, provide a good example of significant advances in this area. Also, in order to realize the benefits from advancements in understanding, it will be necessary to periodically update existing integrated models to ensure that they reflect the fresh insights gained from these new ‘tools for discovery’. This point is further elaborated for the subject of integrated modelling challenges in fusion energy science.

The development of diagnostic instruments capable of making high-resolution measurements of electric and magnetic fields and cross-sectional measurements of turbulent fluctuations has made it increasingly feasible to demonstrate more in-depth correlations between experimental results and theoretical models. This has the potential to greatly improve the basic understanding of the mechanisms controlling plasma confinement. As in dealing with the output from terascale simulations, maximizing the effectiveness of such simulation/experiment comparisons will also necessitate addressing critical computer science and enabling technology (CSET) issues in the area of data management and visualization. Effective utilization of the power of advanced computing to solve challenging problems can be best exploited when the necessary infrastructure is established and effective software tools are made available. Terascale computing requires complementary software that scales as well as the hardware and which provides an efficient code development environment. In general, improved networking, data management and visualization are needed to strengthen the coupling of terascale simulations with theory and experiment.