The operational model of a software system based on the GERT-network

The study proposes an original approach to estimate the reliability of software using the GERT-network model. The approach allows simulating the reliability of software complexes consisting of several interacting software components. As the initial data for reliability evaluation, the values of individual blocks’ reliability estimates are used. Such estima- tions can be assigned by the expert, but also can be obtained as the result of the investigation of the software complex. In this paper, experimental studies were carried out for the “Secure Data Exchange Protocol” software system. Statis- tical estimates of the reliability for the individual program blocks operating were obtained. Using the proposed model, the overall reliability of the entire software package was estimated. The article also proposes an approach to modeling a reliable software architecture based on the idea of multiver- sion programming. The article considers two different ways to implement the multiversion of NVP and RB. The problem of choosing a reliable architecture is formulated in the form of a multi-objective mixed-typed optimiza- tion problem with algorithmically defined objective functions. The criteria to the problem are the overall availability of the software complex and its complexity, which also depends on the number and composition of software components. The problem is solved using a multi-objective genetic algorithm. In the study, various approaches to solving multi- objective optimization problems were considered. A genetic algorithm with a variable length of chromosomes was implemented, which allows encoding software architectures that differ in the number and composition of components. As a result of the genetic algorithm application, various versions of software architectures of the software complex were obtained, which have better reliability. At the same time, the algorithm has proposed to implement multiply ver- sions only for those software components that were not sufficiently reliable.