Grid Enabling of Nano-Science Applications in NAREGI
Grid-Enabling Team(WP6) in National Research Grid Initiative(NAREGI) 1 is developing application-specific middleware components to grid-enable large-scale nano-science applications, Chemistry applications, including those that require coupling of multiple applications on the grid. One example of such applications is multi-scale and/or multi-physics simulation, where each application component utilizes mathematically and physically different modeling and cooperates on spatially or temporally different calculations. To advance such multi-scale and/or multi-physics applications, users have made an every kind of efforts in developing custom codes and decomposing original codes for semantic-level communication between heterogeneous scientific application components. To facilitate easier usage and minimize customization processes of original user programs which may be legacy codes, we have been developing a middleware system, called a _Mediator_ on top of the GridMPI2, that provides high-level transparency in automatically transferring and transforming data between heterogeneous application components. The Mediator focuses on a data-handling specification that correlates different discrete points in finite difference method (FDM), finite element method (FEM), or particle simulations such as Molecular Dynamics(MD) in the unified way. It supports a variety of techniques for semantically transforming the values associated with the correlated points, e.g., in-sphere, first nearest neighbors, and nearest points. The Mediator provides three types of Application Programming Interface (API), which (1) manage a task identification and construct an association between Mediator and application processes in parallel programming style such as Single Program Multiple Data and Master-Worker, (2) register different levels of discrete points, search the correlated discrete points, and determine processes and (3) transfer messages incorporating the extraction and the transformation of the values associated with the correlated points. The prototype system has been applied to multi-scale simulations in nano-science, in which RISM (Reference Site Model) and FMO (Fragment Molecular Orbital) are coupled to analyze an entire electric structure of large-scale molecules immersed in infinite solvent. RISM is employed to analyze the pair correlation functions of molecular sites between a solvent and a solute, while FMO is used to calculate the total electronic energy and the molecular structure of the solute. The interoperability between nano-science applications on the grid might require the functionality to reuse application codes and data from one application domain to another as well as to retrieve and transport data. In the execution step of nano-applications, we have to focus on providing the grid-ready environment which could easily enable the execution, linkage, and coordination of the application modules and the data. To achieve this end, Grid Application Environment Team(WP3) in NAREGI has been developing Workflow Tool, GridPSE and Grid Visualization Tools in cooperation with the resource management mechanism. We expect that these middlewares and tools could be a key component for enabling gateway developers to manage and provide applications on the computational resources for execution and for analyzing linked data sets from related domains, such as Monte Carlo calculations, molecular dynamics, electronic structure studies, and further cross-disciplinary data mining.