Targeted on-line data extraction with SystemXtract

Abstract

Complex system-leve simulation can produce large amounts of data, of which only portions may be of interest. When experimenting with hybrid prototypes, consisting of physi-cal and simulated components, data logs are generated and inspected in real-time. Storing full data logs would not only require much disk space, it would also require much effort to find special events and related system actions afterwards. Targeted on-line data extraction helps to instantly provide the data of interest. We present SystemXtract, a power-ful specification language and tool for on-line data extrac-tion, supporting origin-, value- and dynamic phase-based constrains on the data. The main contribution of this paper is to show how the language can be mapped to a graph-based processing architecture for executing data extraction as specified. Experiments show that the tool-induced over-head in computation time is insignificant and that the real-time execution of the hybrid prototype is not compromised, while the output is reduced to the interesting data.