General Purpose Database Approach for multi-physics Applications
T. Uchdorf and S. Boeck
Introduction
With the steadily increasing computer power scientists in the field of computational materials science are able to perform more calculations with multiple methods / codes on multiple scales (e.g. Ref. [1]). In particular, when investigating trends of material classes the scientists often find themselves confronted with a flood of data with respect to both size and amount of data sets. Furthermore, due to the multi-scale nature of the calculations which are performed, input and output data change in format and units. With this amount of information it becomes hard to keep track of the performed calculations and results as well as putting them into the correct context.
Beside the problem of organizing the data when investigating material trends an information exchange becomes necessary. Therefore, a solution which can automatically keep track of performed calculation in combination with a sophisticated search engine is required for the on-going projects.
Database mapping
In the Algorithm Design and Modeling group a general purpose database approach for multi-physics applications [2] has been developed. In contrast to conventional database applications, changes in the data set structure do not require modifications of the underlying database design anymore. We use the flexible workflow mechanisms available in PHInaX [3] to creating proper database relations fully automatically. The userports of PHInaX modules are automatically mapped to the database table structure.
Security
A security layer can be seamlessly incorporated into the existing general purpose database approach. Therefore, every query is being verified with respect totypical vulnerabilities. Furthermore, all database tables and entries provide both a UNIX-like permission management as well as a modified access control list structure (ACLs).
Results
- Fig.1: Benchmark results up to 1.000.000 randomized database entries
The algorithm to generate general purpose databases is sufficiently fast to provide interactive responds times to the user queries even for large data sets. In Fig. 1 the performance benchmark results for a data set consisting of 1.000.000 randomized data records are depicted. The benchmark has been performed on a standard PC and a MySQL database server. The response time for searching a specific entry out of 1.000.000 data sets is approx. 0.7 sec. Hence, the abstraction penalty for providing a flexible database design is small enough for realistic applications.
Bibliography
- S. Boeck, J. Neugebauer et al., "The S/PHI/nX framework", to be published
www.sphinxlib.de - T. Uchdorf, S. Boeck, J. Neugebauer, “General purpose databases for multi-physics applications”, VDM Verlag, Saarbruecken, 2009, in press
- www.phinax.de
Development of Efficient Render Techniques for Interactive Scientific Visualization
Accomplished Projects
