Crash-proof systemic computing: a demonstration of native fault-tolerance and self-maintenance
Erwan Le Martelot, Peter J. Bentley, R. Beau Lotto · 2008
Reliability in computer or engineering systems is undoubtedly a key requirement in the development process. Safety within critical control systems, and reliable data transfers, require tolerance to unexpected and unwanted phenomena. In biology, new cells can replace damaged cells [1], DNA is able to repair and replicate with error control [1]. These processes are essential to maintain the overall organism. Biology has often been a successful inspiration in computation (artificial neural networks, genetic algorithms, ant colony optimisation, etc) although conventional computation differs widely from natural computation. In this respect, [2] introduced systemic computation (SC), a model of interacting systems with natural characteristics and suggested a new computer architecture. Following this work, [3] introduced a systemic computer as a virtual machine running on conventional computers. In this paper we show, using a genetic algorithm implementation running on this platform, how crash-proof programs following the SC paradigm have native fault-tolerance and easily integrated self-maintenance. KEY WORDS Distributed and parallel computing and systems, systemic computation, fault-tolerance, self-maintenance, crashproof, software reliability. 1