A high level I/O library for numerical simulation data

Abstract

In many research fields of numerical simulations, programs often produce a large amount of mesh data with complex structure. It is a fatal bottleneck for scientists to manage such large-scale simulation data. In this paper, we present a data model with high semantic for simulation data organization, which is based on the data characters such as multi-dimension, multivariable and timespace correlative. We use meta-data to efficiently organize and manage mesh data. proposes the way to recognize characteristics of many kinds of datasets, and give the definition and cataloguing principles for the metadata in scientific computation. Based on the high level data model and metadata, we develop an I/O library for various applications such as simulation and visualization applications or other data analysis tools to access data files with a unified method. Under the support of the I/O library, application programs store simulation data into file systems efficiently with the uniform data format including compressed format, on the other hand, they access datasets with metadata of high semantic, which are described by the data model in numerical simulation fields.