Ùøóññøøø Òòööøøóò Óó Ôòòòò Blockiný Ñóð× Óö ××öúú Ññòòòòññòø

viewpoint real−world viewpoint (1) select services (2) choose matching service−realizations (3) install means of data collection (4) collect data apply management to models (6) find dependencies (7) derive abstract model real− world model abstract model (5) merge datamodel (5) merge data Figure 3: Steps of the Modelling Pro ess 1. First, all servi es and omponents that should be part of the model must be sele ted. Of ourse this depends on the management tasks the model will be used for. 2. For ea h abstra t servi e the mat hing implementations in the real world must be hosen. These may be a number of di erent software implementations, hardware omponents, et . In some ases, as explained in se tion 3, several of these obje ts may nally represent only one instan e of a servi e. 3. Now, the ne essary means of data olle tion may be installed per obje t, using e.g. lo al management agents or tools ommonly available. 4. Then, the real world data may be olle ted, 5. merged a ording to the redu ed model, and 6. dependen ies may be al ulated. This step is the a tual hallenge for the automation using neural te hnologies, as des ribed in the next se tion. 7. If ne essary for the management task, also an abstra t servi e model an be derived, from the model above, together with the information from step 2. It is then possible to apply the required management tasks, e.g. as des ribed in se tion 2, to the models. The rst step is manual, but this is not a real problem, be ause it is dire tly driven by the needs of the management. If only models of the real world dependen ies are needed, the rst two steps may even be omitted; instead, the servi e implementations must then be sele ted dire tly. The installation of the means of data olle tion should | together with the pro ess of olle tion itself | be tied to already used management tools, like management platforms, simple SNMP ([1℄) querying software et . (for a omplete overview see [8℄). Thus, the e ort that must be put into these steps is a eptable. The most omplex task is the nal reation of the model (step 6). As it has to deal with even more obje ts (nodes), it is even more omplex than the original task of dire tly onstru ting the abstra t models. However, with data available from step 5 it is now possible to automate this step. This is des ribed in more detail in the following se tion. 6 Figure 4 also shows possible extensions of the pro ess (dotted lines, marked with asterisks) enabling even more appli ations of the dependen y models. (4) collect data (6) find dependencies (7) derive abstract model real− world model abstract model ** re−instantiate model ** instan− tiated model (5) merge data *