Summary of Requirements

The following list of requirements summarizes the investigations of the overall modeling process and the analysis of the current approaches. Both clearly indicate the need for a new automated modeling approach that:

1.

is able to sufficiently close the lack of dependency information;

2.

handles the scalability issues required by large environments, regarding both model visualization and efforts to create the models;

3.

supports domain concepts (actually as a consequence of the previous requirement);

4.

allows to create models reflecting the actual system behavior in a certain time interval to enable management applications to evaluate short-term changes in the environment as described above;

5.

cleanly decouples data collection (step 3-i) from the actual model creation (step 3-ii) to be able to reuse collected data for (overlapping) models created for different places, e.g., for different departments;

6.

is based on data available for most types of managed resources, but still

7.

needs very little effort to be installed and used in heterogeneous environments (this concerns both the heterogeneity of the probed objects as well as of the runtime-environments for the probes themselves);

8.

has little impact on the performance of the managed systems and networks.

Further concerns like security issues are not covered by this paper; in our prototype security features like authentication of agents and encrypted communication are automatically provided by the agent system we use (the agent based architecture is presented in subsection ).