Universidad Autónoma de Yucatán, Facultad de Matemáticas, Mérida, México CNRS, LAAS, Toulouse, France

---

Abstract—The growing complexity and scale of systems implies challenges to include Autonomic Computing capabilities that help maintaining or improving the performance, availability and reliability of nowadays systems. In dynamic environments, the systems have to deal with changing conditions and requirements; thereby the autonomic features need a better technique to analyze and diagnose problems, and learn about the functioning conditions of the managed system. In the medical diagnostic area, the tests have included statistical and probabilistic models to aid and improve the results and select better medical treatments. We propose a probabilistic approach to implement an analysis process. The base of our approach is building a Bayesian network as model representing runtime properties of the Managed Element and their relationships. The Bayesian network is initially built from monitored data of an Enterprise Service Bus platform under different workload conditions, by means a structure learning algorithm. We aim to improve the functionalities of an Enterprise Service Bus platform integrating monitoring and fault diagnosis capabilities. A case study is presented to prove the effectiveness of our approach.

Index Terms—Autonomic computing, bayesian network, probabilistic reasoning, diagnostic, machine learning, SOA

[PDF]

Cite: Roberto Koh-Dzul, Mariano Vargas-Santiago, Codé Diop, Ernesto Exposito, Francisco Moo-Mena, Jorge Gómez-Montalvo, "Improving ESB Capabilities through Diagnosis Based on Bayesian Networks and Machine Learning," Journal of Software vol. 9, no. 8, pp. 2206-2211, 2014.