Service-Oriented Autonomic Computing Based On PDC-Agent

Currently, it has been a research hotspot in the areas of distributed systems to enable the process of sharing and integration of heterogeneous web applications to realize self-management (autonomic computing), with the guidance of dynamic business requirements and strategies.This paper firstly reviewed the dominant technologies for dynamic compositon of heterogeneous web applications, their development trends, as well as the concept of autonomic computing and its state of the art, and then analyzed the main problem about applying current technologies to establishing autonomic computing systems, i.e., how to express high-level business requirements and stratigies and integrate them into the decision-making process of the intelligent agents, to make agents be aware of the variations of high-level business requirements and strategies when they are conducting autonomous resource composition. This paper argued that the primary reason of this problem is the inability of traditional agent model that can't reflect the variations of high-level business requirements and strategies. In the traditional agent model, the management logic of the systems is integrated with the application logic. They are encoded when system is implemented and can't be modified dynamically. This paper coped with this problem by separating management logic from application logic and integrating policy-based management method into multi-agent model. On the one hand, the management logic of system is expressed as policies. Since policies can be dynamically added, updated and deleted, they provide an efficient approach to reflect dynamic requirements and stratigies. On the other hand, by integrating policy method into agent mental model, agents can be guided by high-level business strategies when they are making decisions.On the basis of the above ideas, this paper proposed a model of service-oriented autonomic computing system based on PDC-agents. This model is founded on the service-oriented architecture and its supporting technologies. So, in this model, resources are abstracted as services, which are the management objects of agents. As autonomic elements, agents take part in cooperation by the guidance of high-level business stratigies, to realize service discovery, negotiation, compositon, authentication and authorization, process monitoring, and exception treating, etc. Cooperation relationships are established and operationalized by means of contracts, which make the cooperation among autonomic elements more stable and more reliable. The key parts of this modelinclude the mechanism and theory of policy description, refinement and deployment, and the agent model that is suitable for acting as an autonomic element.Although policy-based methods have attracted great attentions, and a number of achievements have been made, how to apply this method to guiding the behaviors of agents is still a new rearch area. As current policy descripton languages can't be combined with traditional BDI logic for establishing an extended agent model, this paper proposed a policy description language based on concept instance patterns, as well as a mechanism for policy refinement and deployment.Meanwhile, traditional BDI-based agents only concern with their own desires, and can't reflect external motivations, so they are unsuitable for acting autonomic elements of autonomic computing systems. This paper proposed a policy and contract extended BDI logic, called BGIpdc logic, by integrating contracts and policies into traditional BDI logic. In this logic, the motivations of agent are composed of internal motivations (its own desires) and external motivations that are brought about by policies and contracts. These various kinds of motivations are taken as goals of PDC-agent after conflicting treatment. The agent based on BGIpdc logic is called PDC-agent, whose architecture and working principle are formulated in this paper.Finally, this paper further analyzed the implementation aspect of PDC-agent-based service-oriented autonomic computing system by using a case study.