Hybrid Inference Architecture and Model for Self-healing System

Abstract

Distributed computing systems are continuously increasing in complexity and cost of managing, and system management tasks require significantly higher levels of autonomic management. In distributed computing, system management is changing from a conventional central administration, to autonomic computing. However, most existing research focuses on healing after a problem has already occurred. In order to solve this problem, an inference model is required to recognize operating environments and predict error occurrence. In this paper, we proposed a hybrid inference model – ID3, Fuzzy Logic, FNN and Bayesian Network – through four algorithms supporting self-healing in autonomic computing. This inference model adopts a selective healing model, according to system situations for self-diagnosing and prediction of problems using four algorithms. Therefore, correction of error prediction becomes possible. In this paper, a hybrid inference model is adopted to evaluate the proposed model in a self-healing system. In addition, inference is compared with existing research and the effectiveness is demonstrated by experiment.

This work was supported in parts by Ubiquitous Autonomic Computing and Network Project, 21th Century Frontier R&D Program, MIC, Korea, ITRC IITA-2005-(C1090-0501-0019), Grant No. R01-2006-000-10954-0, Basic Research Program of the Science & Engineering Foundation, and the Post-BK21 Project.