Модель надежности многопроцессорных аппаратно-программных комплексов систем управления реального времени с мультиверсионным программным обеспечением

Abstract

Reliability is a critical parameter of real-time control systems. In practice, the reliability of hardware and software complexes included in such systems, is ensured by creating redundancy of hardware components and a multi-version approach to software development. But the redundant reservation of hardware devices and too many versions of software can lead to unjustified growth in the cost of creating and operating the projected control system. A rational approach to the design requires the creation of a model, which allows evaluating the reliability of different hardware and software complex configurations at the design stage. We proposed a mathematical reliability model of hardware and software complexes of real-time control systems, built with the use of mathematical apparatus of the queueing theory, which is a system of differential equations for the probability of finding states in the system, in which one or another component of the hardware and software complex is faulty. From the system of differential equations, the system of linear algebraic equations for probabilities of states in a steady-state mode was obtained. An analytical solution of this system is given, which allows us to evaluate the reliability of multiprocessor hardware-software complexes with multi-version software without significant expenditure of computational resources. Possibilities of using the results obtained to optimize the reliability of multiprocessor hardware-software complexes with multi-version software are presented and directions for further research are proposed