Development of a common data model for scientific simulations
John J. Ambrosiano, Butler, D. M., Celeste Matarazzo, Mark D. Miller, L.A. Schoof · University of North Texas Digital Library (University of North Texas) · 1999
The problem of sharing data among scientific simulation models is a difficult and persistent one. Computational scientists employ an enormous variety of discrete approximations in modeling physical processes on computers. Problems occur when models based on different representations are required to exchange data with one another, or with some other software package. Within the DOE`s Accelerated Strategic Computing Initiative (ASCI), a cross-disciplinary group called the Data Models and Formats (DMF) group, has been working to develop a common data model. The current model is comprised of several layers of increasing semantic complexity. One of these layers is an abstract model based on set theory and topology called the fiber bundle kernel (FBK). This layer provides the flexibility needed to describe a wide range of mesh-approximated functions as well as other entities. This paper briefly describes the ASCI common data model, its mathematical basis, and ASCI prototype development. These prototypes include an object-oriented data management library developed at Los Alamos called the Common Data Model Library or CDMlib, the Vector Bundle API from the Lawrence Livermore Laboratory, and the DMF API from Sandia National Laboratory.