Generation of synthetic XML for evaluation of hybrid XML systems

Abstract

Hybrid XML storage offers a large number of alternative shredding choices. In order to automatically determine optimal shredding strategies it is crucial to have an insight into how the structure of a XML data set affects the performance. Since the structure can take many forms and the number of possible mappings is huge it is important to gain insights on the relation between structure and performance for formats that are actually used. By taking real-world data sets and modify the structure in steps you can see how the performance and other measurable properties change. We describe how a data generator can be used to produce a synthetic data set based on an existing data set, by using four different models. We compare the performance on the original data set with the performance on the different synthetic models. © 2010 Springer-Verlag Berlin Heidelberg.