The technology of creating a fault-tolerant multi-module software complex based on the procedure of mutual internal checks

Abstract

The article considers the problem of creating a fault-tolerant multi-module software complex based on the procedure of mutual internal checks. This problem is especially acute in cases of use of critical computer systems, where the requirements for software security are extremely high. The main idea of the article is that a multi-module software complex, as a system based on computing resources, both at the stage of creation and at the stages of operation, can independently carry out diagnostics between individual modules, thus ensuring its operability. The purpose of the article is to develop a procedure of mutual internal checks for its implementation in the technology of creating a fault-tolerant multi-module software complex. The diagnostics model is based on the assumptions that the connections between modules can be represented by a fully connected undirected graph and each pair of modules can exchange messages. Each module uses self-diagnostic tools to determine its status "good" or "bad") and transmits it to another module. After accumulating the results of checks, a collective decision is made regarding the serviceability of the modules. An original algorithm for the implementation of diagnosis technology has been developed. The correctness of the developed concept of ensuring resistance to failure of software complexes by diagnostics was verified by mathematical modeling. Graphs of dependences of diagnosis reliability on the number of modules in the program and the number of permissible failures were obtained. Confidence intervals were estimated, which show that as the number of failures increases, the confidence interval expands, and when it decreases, it narrows. It was concluded that the developed algorithm requires less redundancy of the system and allows to obtain the result of only two rounds of message exchange between the program modules. At the same time, it provides diagnostics of a multi-module software complex when almost half of its modules fail.