Maidaniuk I. Methods and tools for supporting statistic experiments executing with GL models on estimating the reliability of fault-tolerant multiprocessor systems

In this thesis the tools that allow efficiency increasing of reliability calculation of fault-tolerant multiprocessor systems (FTMS) are researched and proposed. The reliability calculation is based on the method of statistic experiments with graph-logic models (GL models). Comparative analysis on various parameters demonstrates the benefit of 2r-model usage to calculate the reliability. A modification of the known method, and a new algorithm for forming the basic functions of the 2r-model edges are proposed, the methods mention above have lower complexity. The equations, which enable to prognosis the complexity of the nonbasic FTMS GL-models, before the modification, is proposed. It is proposed autonomous generator of binary vectors of a given weight, based on controllable shift register. The efficiently of using such solution for modelling the system state vector is proved. The structure of a specialized computing system for calculating the reliability of the FTMC, which implements the main results obtained in the thesis is developed.