| In the thesis we view Web Intelligence (WI) from a distributed resource optimization perspective. Specifically, we propose a generalized distributed resource optimization problem for WI, which concerns how to distribute service requests to different resource nodes such that their service request load is approximately balanced. Accordingly; they are utilized in an approximately optimal way. We then provide an Autonomy Oriented Computing (AOC) based approach for the proposed problem. AOC is a multi-entity based methodology for addressing complex systems modeling and computationally hard problem solving. Its essence lies in the idea of self-organization of autonomous entities through interactions. Due to this reason, it is quite suitable to those large-scale, highly distributed problems. In the provided approach, autonomous agents are employed to carry service requests submitted to a resource environment in order to find appropriate resource nodes for them. They indirectly interact with each other via the resource environment (e.g., the joining or leaving of an agent at a resource node influences the decision making of other agents on whether or not joining or leaving this resource node). Through this interaction mechanism, agents may self-organize themselves and distribute service requests to resource nodes to achieve load balancing and resource optimization.;Usually, Web-based resource environments can be classified into two types, namely, homogeneous environments and heterogeneous environments. By fully considering the distinct features of these two types of resource environment, we refine the above proposed AOC-based resource optimization mechanism into two specific ones. For the mechanism suitable to homogeneous environments, we propose mathematical models to macroscopically characterize the process of resource optimization. Through numerical simulations of the models and experiments on a computing platform, called SSADRO; we validate the proposed mechanism. For the one suitable to heterogeneous environments; through experimentation also on the SSADRO computing platform we validate its properties, such as, effectiveness and robustness.;In general, this thesis tries to integrate Web Intelligence (WI) and Autonomy Oriented Computing (AOC): (1) It introduces Autonomy Oriented Computing (AOC) as a general methodology to Web Intelligence (WI) for addressing the issue of distributed Web resource optimization; (2) Through the distributed Web resource optimization problem, it re-validates the effectiveness of AOC in solving real-world, large-scale, highly distributed problems. (Abstract shortened by UMI.)... |
