Using Adaptive Neural Networks in Self-Healing Systems

Self-healing in systems is one of the main characteristics of Autonomic Computing (AC). In this regard the challenge is how to implement self-healing systems in real time, since online learning is required so that the running system is tuned and adapted automatically, based on the current changes of the system´s behavior. In this paper, to overcome the challenges associated with self-healing comprising monitoring, interpretation, resolution, and adaptation (MIRA); a novel technique is implemented using pipelined recursive neural networks (PRNN) with a modification of the original algorithm. This method enables us to deal with several independent signals instead of one input; also on the fly learning is achieved in order to address the problems of providing system self-healing. For example instigating continuous learning rather that supervised learning which is not suitable for real systems. Traditional Predefined methods were used, but in this paper an external approach of self-adaptation is implemented to suit the changes of real time systems.