The Embodiment of Autonomic Computing in the Middleware for Distributed System with Bayesian Networks

摘要

In this paper, we propose an intelligent middleware framework based on the concept of the autonomic computing. Because decisions for optimizing a system are highly dependent on the status of the system in the near future, predictions of the status can boost up the performance of system. Bayesian networks that are frequently used in the domain of user modeling are adopted as a diagnosis engine. Since there are many different kinds of behavioral patterns, it is impossible to model all of them into a single model. A Bayesian network is learned for each typical behavioral pattern. Experimental results show that the proposed method offers a promising way of intelligent middlewares.