Research On Multi-Agent Oriented And Policy-Based Operating Mechanism

Nowadays, the systems are becoming more and more complicated. How to manage these systems effectively has already become a hot issue now. Generally speaking, in order to manage the complicated systems effectively, the distributed systems in particular, two problems need to be resolved: 1) autonomy of the system behavior. How to manage a system by the system itself with much less human efforts to reduce the workload of the administrators greatly. 2) controllability of system management. An approach and mechanism of changing the system behavior is needed to make the system work in an expected way.We begin our research work with focus on the autonomy and controllability of system management and with multi-Agent coordination as the target of our research. At the same time, the policy is introduced to the system management. The police refinement method and the operating mechanism of policy-based management are described in an exhausted way. All of these works bolstered the new exploration to realize a policy-driven based and multi-agent coordination system. a policy refinement method based on concept decomposition has been proposed and the process of policy refinement can be simplified by making use of the ontology which is excellent in descripting semantic relationships. In general, our works and contributions are shown below.a) A policy refinement method based on concept decomposition has been proposed. Supported by the advantages of the ontology in describing semantic relationsips among concepts, through building policy ontologies and related domain ontologies , using ontology semantic reasoning, the medhod realizes the decomposition process of from a restiction which itself is a concept in an ontology to a set of concepts that are semantically unitary. When the business or application domain changes, we only need to modify or replace relevant ontologies to keep the results of policy refinement up-to-date.b) A new policy operating mechanism has been proposed. In our mechanism, policies, managed objects and behaviors of managed objects are considered as three independent components. The policy manages the object. The behavior of a managed object has to be executed through the managed object. Policy is used to control the behavior-selecting process of a managed object. Through policy, the managed object is triggered to switch its behavior to realize the running of a policy. The method makes the running of a policy more simple, the whole operating mechanism can be modified or expanded according to different requirements to suit different situations.