Immune system (IS) is capable of evolving, learning, recognising and eliminating foreign molecules which invade organisms. Fault tolerant approaches consist in detecting erroneous states of an algorithm and executing recovery procedures. They are generally implemented by adding formal properties to be satisfied by state variables during execution. However, defining fault-tolerant procedures can be error-prone and sometimes equivalent to a formal proof. Moreover, it is always very difficult to make a sound assumption about the typology and frequency of real. We propose an analogy between IS and fault tolerant computing. After a brief presentation of immune algorithms and especially lymphocyte simulation (B-cells, suppressor T-cells, etc.) using genetic operators, we present an immune model for detecting and recovering erroneous states during execution. Program states are examined by procedures playing the role of B-cells. Those properties have previously been learnt, automatically, during testing phase using mutation analysis techniques. When a program state has to be recovered, T-cells procedures are activated in order to recover erroneous states. In this way, each software algorithm may develop automatically its own immune system capable of evolving and stimulating specific responses to software failures. We finish by a brief discussion of the main possibilities of an immune approach to fault-tolerant computing.
CITATION STYLE
Xanthakis, S., Karapoulios, S., Pajot, R., & Rozz, A. (1996). Immune system and fault-tolerant computing. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1063, pp. 181–197). Springer Verlag. https://doi.org/10.1007/3-540-61108-8_38
Mendeley helps you to discover research relevant for your work.