Morozov K. Methods and tools for building models of behavior of non basic fault tolerant multiprocessor systems

The thesis is devoted to the problem of building graph-logical models (GL models) of non-basic fault-tolerant multiprocessor systems (FTMS) for calculating their reliability parameters by means of carrying out statistical experiments with the above-mentioned models. A method for transforming of GL-models by modifying the expression of any edge function of so-called MLE-model is proposed. The method is based on changing either any one or both two fragments of the edge function’s expression and allows modifying the model so that on some system state vectors, it starts to show an operable and/or inoperable state in contrast to the original one. The boundaries of influence for general case of modification of both single and several edge functions on change of the behavior of the model is analyzed. For the first time, an analytical apparatus that allows to determine the state vector of t-diagnosable FTMS based on the results of mutual testing of processors, according to the Preparata-Metze-Chen model (PMC-model), for arbitrary topology of connections between them has been proposed. For the first time, a method of building GL-models of so-called weighted systems, in which each component has a certain weight that characterizes its contribution to the performance of the system as a whole as well as a generalization of method of building GL-models, based on the usage of multivalued logic edge functions allowing them to be used as models of systems for which, as well as for their components, more than two states of operability are available are proposed. A method of building GL-models of hierarchical systems by means of composition of several MLE-models as well as a method of building models of systems consisting of several subsystems and having a sliding reserve are proposed.