hide
Free keywords:
-
Abstract:
Advanced applications in a metacomputing environment need to communicate semantic information in addition to raw data. This requires a suitable data model, appropriate concepts on how to exchange data objects, as well as corresponding I/O layers and communication protocols. In this paper a data model is presented, providing a conceptual base for the design of generalized grid operations and semantics aware I/O layers. With this model, code redundancies can be immensely reduced and high reusability of algorithms is ensured. It is inspired by the mathematical theory of fiber bundles and generalized to the concept of index spaces. The presented concepts allow us to formulate data properties and relationships among data objects in a powerful and widely usable way. A wide range of data types, like 3D uniform grids, particle trajectories, triangular surfaces, expansion coefficients, can be covered. The HDF 5 API and XML can be used to realize fully capable I/O layers. By employing HDF 5 virtual file drivers remote I/O of persistent and volatile data is easily achieved, in addition to local file I/O.