[1.7ex] Munich Network Management Team
University of Munich, Dept. of CS
D-80538 Munich, Germany
Phone: +49-89-2178-2171, Fax: -2262
Various service management tasks depend on knowledge about objects they are applied to. Understanding dependencies between these objects is of special interest. Such dependency models are neither generally available nor easy to create. Creating them manually is expensive in regard to both time and money. In addition, they must be brought up to date, reflecting the frequent changes in the managed environment.
After motivating use of dependency models for various service management tasks, this article presents a new approach on the modelling itself. It is shown that the proposed modelling process enables automated generation of such models. To add flexibility and to guarantee applicability in many environments a special technique based on neural technologies is used. This avoids or at least reduces the disadvantages of a purely manual creation process.