In principal, the applications are similar to the ones described in the previous subsection, with the difference that abstract models are:
However, the models' restriction to the abstract level of course has implications on their usability. This problem is typically circumvented by one of two methods: Virtual instantiation, keeps lists of real objects (together with the object specific management information) for each class, but the inference engine's main work is carried out on the abstract models. Full instantiation maps the abstract to environmental models. In this case the former are used to add commonly known dependency information to specific scenarios, but after their instantiation they are not directly operated on by the management tools.
More direct use of abstract models has been carried out in several Model Based Reasoning (MBR, [#!pemt95!#]) systems, for a number of years. These have already been able to map the results of errors in simple components to services or systems visible to end users. Another application is to diagnose possible sources of errors on lower layers, if problems on higher ones are reported. Still, management tools based on abstract models so far have not been very successful with regard to their market share. The number of available models remained too small to let the strengths of the tools become fully visible: On the one hand, companies delivering products do not want to enclose models due to the extra efforts needed and because of strategic policies enforced to protect the companies' market positions (e.g., confidentiality). On the other hand, it is also virtually impossible for the providers of the management tools to supply sufficient models themselves.
As a partial solution to this problem a standardized and widely accepted library for the most important classes could be used. In the past, efforts to do so have not been very successful. However, the endeavors of the Distributed Management Task Force (DMTF) for the Common Information Model (CIM, [#!cim22!#]), where esp. the Common Schemas are able to serve as a basis for further abstract models, hopefully will be more successful. The following subsection provides more information on models similar to CIM.