Root Cause Analysis and Proactive Problem Prediction for Self-Healing

As the rapid evolvement of distributed computing system, the requirements imposed on problem determination techniques are increased to help system control and manage in high levels of automated ways, which represents the capability of self-healing. Many artificial intelligent approaches are widely used in the fields of fault managements. In this paper, we propose an approach to fault management for self-healing system through learning and analyzing real-time information, to provide both root cause analysis and proactive problem prediction. Using Bayesian network algorithm, we describe a complex system as a compact model that presents probabilistic dependency relationships between various factors in such a domain. We also provide an improved process that deals with collected parameters in advance, which enhances learning efficiency and reduces learning time. For estimating the efficiency and accuracy, an experimental demonstration based on system performance measurements is implemented and evaluated via diverse comparisons, which shows the availability is optimistic.