Implementing Unified Access to Scientific Data from.NET Platform

Scientific data differ from common relational data in many aspects: scientific data may have a very complex structure, they are usually stored in files of various formats and individual data items can be very large. In this paper we present an extensible and efficient client-server system for accessing scientific data and its metadata. The architecture and major capabilities of our system will be described in the paper. The core of our approach is an extensible XML-based structure that annotates scientific data with rich metadata and maps every file or part of a file to a named strongly typed entity. We do not introduce any new file formats and file transfer techniques, thus our approach doesn’t require major changes to existing computational software. SOAP protocol and Web Services are used for accessing data sets and performing data requests. Filtering and caching enables an efficient access to large portions of data over network. Example of implemented filters are cropping and thinning of 2D and 3D arrays. Our system is fully extensible and allows adding new data types, new file formats and new filtering algorithms without changing its core algorithms. Now it is used for accessing results of computational fluid dynamics simulations, but we hope that it can be adapted to many branches of science. The client is implemented on the.NET platform; the server-side is currently running on the IBM Regatta SMP mainframe on AI