Analysis and identification of network failures in the information network

Abstract

Listen

Translate article

This article shows that each information system has its own characteristics, which are determined by the scope of its application. The importance and responsibility of tasks solved using real-time systems have led to high requirements for the reliability of these systems, and the failure of the entire information system or its individual components can lead to negative consequences. Network failure recognition is based on the principle of determining the nature of network failures based on the product system. This principle is implemented through a dialogue procedure within the framework of the Bayesian approach, which allows you to accumulate information coming from various sources in order to confirm (not confirm) a certain hypothesis. A logical conclusion management strategy was developed, the control parameters of which are the current probability of the truth of the hypotheses, the limits of their change, and the weight of the signs associated with these hypotheses. Accounting for the current probability of hypotheses and the limits of their change allows in the process of conclusion, firstly, to focus attention on the most promising (probable) hypothesis, and secondly, to stop the conclusion after reaching the upper threshold, which will reduce the number of signs to be checked, and thus shorten the dialogue time yourself. In addition, having reached the lower threshold, the hypothesis is rejected as implausible and no longer participates in the conclusion process, which also leads to a reduction in recognition time. Accounting for the weights of signs in the process of logical inference allows you to first of all check those signs that maximize the probability of the plausibility of hypotheses. In general, the developed strategy, unlike the classical scheme, generates a purposeful process of testing the plausibility of hypotheses, which leads to a reduction in recognition time.