Learning fault-tolerance from nature

In the last decade, there has been a considerable increase of interest in fault-tolerant computing due to dependability problems related to process scaling, embedded software, and ubiquitous computing. In this paper, we discuss an approach to fault-tolerance which is inspired by biological systems. Biological systems are capable of maintaining their functionality under a variety of genetic changes and external perturbations. They have natural self-healing, self-maintaining, self-replicating and self-assembling mechanisms. We present experimental and numerical evidence that the intrinsic fault-tolerance of biological systems is due to the dynamical phase in which the gene regulatory network operates. The dynamical phase is, in turn, determined by the subtle way in which redundancy is allocated in the network. By understanding the principles of redundancy allocation at the genetic level, we may find ways to build chips that possess the inherent fault-tolerance of biological systems.