Policy conflict analysis in distributed system management

Distributed system management is concerned with the tasks needed to ensure that large distributed systems can function in accordance with the objectives of their users. These objectives are typically set out in the form of policies that are interpreted by the system managers. There are benefits to be gained by providing automated support for human managers, or actually automating routine management tasks. To do this, it is desirable to have a model of policies as objects that can be interpreted by the system itself. The model is summarized. It is clear that there is the potential for conflicts between policies. These conflicts may be resolved informally by human managers, but if an automated system is to recognize them and resolve them appropriately, first it is necessary to analyze the types of conflict that may occur. We analyze the types of overlap that may occur between policies, and show that this analysis corresponds to several familiar types of policy conflict. Some possible approaches to the prevention and resolution of conflicts are suggested, and this work is put into the context of other work on policies and related areas, including deontic logic.